VOLUME 15, ISSUE 1, JANUARY 2026

Abstract: Women’s safety has become a critical concern due to the rising number of harassment and assault incidents, particularly in public and isolated areas. Conventional safety measures such as mobile applications and emergency helplines often fail to provide immediate assistance during panic situations. This paper presents the design and implementation of a Smart Footwear System for Women Safety, an IoT-based wearable solution that ensures continuous availability, discreet operation, and rapid emergency response. The proposed system integrates a microcontroller, pressure or motion sensors, a GPS module, and a GSM communication unit embedded within footwear. The system can be activated either manually using a concealed switch or automatically through abnormal sensor activity. Upon activation, an emergency alert containing real-time location coordinates is transmitted to predefined contacts via GSM, while an audible buzzer is triggered to attract nearby attention. The system is designed to be compact, low-power, and comfortable for daily use without affecting normal walking behavior. Experimental results demonstrate reliable alert transmission, accurate location tracking, and effective hardware–software integration. This wearable safety solution highlights the potential of IoT-enabled assistive technologies in addressing real-world societal challenges.

Keywords: Women Safety, Smart Footwear, IoT, GPS, GSM, Wearable Devices.

Abstract: The exponential growth of wireless data traffic and the increasing limitations of radio frequency (RF) spectrum have driven the need for alternative high-bandwidth and interference-free communication technologies. Light-Fidelity (Li-Fi), a subset of Visible Light Communication (VLC), has emerged as a promising solution by utilizing light-emitting diodes (LEDs) for high-speed wireless data transmission. This paper presents the design and implementation of a Li-Fi-based system capable of transmitting text, audio, and image data using intensity modulation of visible light. The proposed system employs microcontroller-based encoding at the transmitter to modulate LED light signals according to the input data, while a photodetector-based receiver decodes the transmitted information using signal conditioning and amplification circuits. Text data is transmitted using serial communication protocols, audio signals are conveyed through real-time analog modulation, and image data is transferred as sequential binary streams and reconstructed at the receiver end. Experimental results demonstrate reliable short-range indoor communication with minimal electromagnetic interference and improved data security compared to conventional RF-based systems. The study validates Li-Fi as a cost-effective, energy-efficient, and secure wireless communication technology suitable for multimedia data transmission in environments where RF communication is constrained. The proposed system highlights the potential of Li-Fi for future applications in smart indoor communication, healthcare, educational institutions, and secure data transmission systems.

Keywords: Li-Fi Technology, Optical Wireless Communication, LED Modulation, ESP32, Text Transmission, Audio Transmission, Image Transmission.

Abstract: This study introduces Hydrosan, a smart autonomous robot designed for river detoxification to combat widespread water pollution. The robotic system is equipped with a dedicated mechanism that effectively collects floating plastics and other solid waste from the surface of rivers. It also incorporates multiple sensors to continuously assess water quality parameters, such as pH, in real time. Autonomous operation is achieved through an intelligent control architecture supported by GPS-based navigation. Experimental field evaluations demonstrate strong performance in both waste removal and environmental data monitoring, highlighting Hydrosan as an eco-friendly, automated solution with significant potential for large-scale river cleaning and environmental protection initiatives.

Keywords: HydroSan, River Detoxification, Water Pollution Control, Floating Waste Removal, Autonomous Cleaning Robot, ESP32, IoT-Based Monitoring.

Abstract: Timely detection and response to cyber threats have become increasingly challenging due to the distributed and dynamic nature of modern digital infrastructures. Conventional centralized intrusion detection systems struggle with scalability, delayed response, and privacy concerns when handling large volumes of security data. This project proposes an Adaptive Federated Threat Detection framework that employs federated learning to collaboratively identify malicious activities across multiple decentralized nodes without sharing raw data. Each participating node trains a local threat detection model using its own security logs, while a central coordinator aggregates encrypted model updates to build a global intelligence model. The system continuously adapts to evolving attack patterns by refining detection strategies based on real-time feedback. Experimental evaluation demonstrates improved detection accuracy, reduced false alarm rates, and enhanced robustness against emerging threats compared to traditional centralized security approaches, while preserving data confidentiality.

Keywords: Federated Learning, Adaptive Threat Detection, Cybersecurity, Distributed Intrusion Detection, Privacy-Preserving Machine Learning, Intelligent Security Systems

Abstract: The proposed system aims to create a safe online voting system by developing a face recognition-based authentication model using K-Nearest Neighbors (KNN). The system verifies voters through unique credentials and a real-time photo captured via webcam. It strictly enforces a one-vote-per-user policy by comparing live images with stored database records. Results are updated every minute, ensuring transparency and efficiency while reducing long queues at polling stations.

Keywords: Face recognition, KNN, Online Voting System, Machine Learning.

Abstract: Recent improvements in AI-driven content synthesis have made it possible to fabricate highly realistic human imagery and video sequences. These artificially produced visuals now resemble genuine recordings so closely that older inspection practices struggle to recognize what has been digitally altered. This situation has amplified concerns related to online authenticity, public trust, and misuse of sensitive digital content. Because manual or traditional forensic checks no longer keep pace with advanced manipulation tools, dependable automated detection systems have become essential. Models built using convolution-based learning strategies are frequently employed for this purpose because of their ability to uncover fine-grained visual abnormalities introduced during fabrication. This article reviews studies exploring such models for identifying altered still images, detecting AI-produced visuals, and analyzing videos using both spatial patterns and motion information. While these approaches show promising performance on curated datasets, they often face difficulties under real-world conditions. The review highlights current limitations—such as softness under compression, dataset dependency, and weak generalization—and identifies the need for stronger multimodal and adversarial-resistant solutions.

Keywords: Deepfake Detection, Convolutional Neural Networks, AI-Generated Images, Video Forgery Analysis, Multimedia Forensics, Digital Authenticity

Abstract: Document verification is an essential process in academic, governmental, and corporate sectors. However, traditional document verification systems are time consuming, prone to forgery, and lack transparency. This paper proposes a Blockchain- Based Document Verification System that ensures secure, tamper-proof, and transparent document storage and validation using blockchain technology. The system integrates web technologies for user interaction, a backend for processing, and blockchain for immutable verification. This solution aims to eliminate document fraud, reduce verification time, and increase trust between organizations and individuals.

Keywords: Blockchain, Document Verification, Smart Contracts, Ethereum, Web Application, Security, Transparency.

Abstract: Road accidents constitute a leading cause of mortality and injury globally, necessitating rapid detection and emergency response mechanisms. This paper presents the design and implementation of an Internet of Things (IoT)-based vehicle accident detection and automated emergency notification system. The proposed system integrates an MPU6050 accelerometer- gyroscope module and NEO-6 GPS receiver with an ESP32 microcontroller to enable real-time impact detection and precise location tracking. Upon detecting abnormal deceleration or collision events, the system activates an audible alert, allowing occupants to confirm their safety via a manual button. If unacknowledged within a predefined timeframe, the system automatically transmits GPS coordinates to a web-based monitoring platform using secure tunneling protocols. The implementation demonstrates reliable accident detection with minimal false positives, real-time location mapping, and automated emergency notification capabilities. Experimental results validate the systems effectiveness in reducing emergency response time while maintaining a compact, power-efficient design suitable for diverse vehicular applications. This holistic approach leverages IoT connectivity to bridge the gap between traditional passive safety features and modern emergency management systems, ultimately contributing to enhanced road safety and reduced fatalities. The system employs sophisticated threshold-based algorithms to distinguish genuine collision events from normal driving scenarios such as sudden braking or speed bumps, achieving a true positive detection rate of 94.7 percent while maintaining a false positive rate below 2.3 percent. Power management is optimized through dual TP4056 charging modules managing parallel connected 3.7V lithium on batteries, enabling over 12 hours of continuous operation. The embedded firmware, developed using ESP-IDF framework, implements multi-threaded processing for concurrent sensor data acquisition, collision analysis, and wireless communication tasks. The web-based visualization platform utilizes Leaflet.js mapping library to provide dynamic, real-time tracking of vehicle location with sub-second update latency. Ngrok tunneling facilitates secure embedded-to-web communication, allowing the ESP32 to transmit incident data without requiring static IP configuration or complex network setup. System validation through controlled laboratory experiments and field testing confirms detection latency of less than 180 milliseconds from impact to alert activation, and end-to-end notification delivery within 600 milliseconds. The modular architecture supports easy integration with existing vehicles as a retrofit solution, requiring minimal installation effort and no modifications to factory-installed systems. This research contributes to intelligent transportation systems by demonstrating a scalable, cost-effective approach to vehicular safety that operates independently of external infrastructure. The proposed solution addresses critical limitations in current accident response mechanisms, potentially reducing emergency response time by up to 40

Keywords: IoT, accident detection, ESP32, MPU6050, GPS tracking, emergency notification, embedded systems, real-time monitoring

Abstract: This paper presents a novel Neurosymbolic AI framework designed to enhance the accuracy and explainability of brain tumour diagnosis. By combining deep learning architectures (VGG16 for classification and U-Net for segmentation) with a symbolic genomic rule engine, the system integrates structural MRI data with molecular biomarkers such as IDH mutation and MGMT promoter methylation status. This multi-modal approach achieves high-fidelity risk assessments while providing clinicians with "white-box" explainability through Grad-CAM heatmaps and guideline-based treatment recommendations. Additionally, the system features an interactive Student Learning Lab and a federated learning hub to support decentralised training and medical education.

Keywords: Neurosymbolic AI, Brain Tumour Diagnosis, Explainable AI (XAI), Multi-modal Fusion, Deep Learning, Genomic Reasoning.

Abstract: ELARA is a lightweight outfit recommendation system that combines explainable styling rules with basic image processing. It creates embeddings from clothing images using MobileNetV2 and compares them to fashion datasets using cosine similarity to determine the type of item. ELARA creates rule-based outfit recommendations based on this match to actual styling principles. Additionally, users can search datasets for suggestions. To determine appropriate and inappropriate color schemes, the system incorporates a webcam-based color analysis tool and a fashion chatbot. All things considered, ELARA provides a quick, convenient and easy-to-use solution for daily fashion advice.

Keywords: MobileNetV2, Image embeddings, Cosine similarity, Chatbot.

Abstract: Athletic performance relies heavily on how the foot makes contact with the ground during movement. Most traditional methods of measuring this involve lab equipment, which can feel unnatural and restrict testing to controlled environments. This project introduces RUCK SOLE X, a smart insole designed to track foot biomechanics during actual training and competition. The insole is flexible, comfortable, and made for long-term use. It combines pressure, motion, temperature, and moisture sensors to gather detailed foot data without interrupting the athlete. An embedded controller processes this information in real time, instantly identifying gait events and load patterns, even without constant internet access. For deeper insights, long-term data is analyzed in the cloud to identify unusual movement patterns and potential injury risks. All results are shown through a mobile app with clear, easy-to-read visuals, helping athletes and coaches quickly grasp what’s happening with their feet. Testing indicates the system delivers reliable performance and fast feedback, working well in real-world training settings. Its adaptable design fits many types of athletic shoes, making it suitable for various sports. Regular firmware updates enhance accuracy over time. The collected data helps coaches and sports scientists refine training plans, lower injury risks, and personalize performance improvements. Overall, RUCK SOLE X represents a significant advance in wearable sports technology.

Keywords: Athlete Monitoring, Biomechanics, Edge Computing, Gait Analysis, Sensor Fusion, Smart Insole, Sports Technology, Wearable IoT

Abstract: This paper proposes an image-to-image interior design generation framework based on latent diffusion modeling. Given a single RGB image of an indoor space, the system formulates interior redesign as a conditional generation problem, where structural geometry is preserved while visual attributes are optimized under style-specific constraints. A Stable Diffusion backbone is employed with controlled text-conditioning and spatially consistent sampling to generate multiple décor configurations. The framework incorporates preprocessing for viewpoint normalization and semantic alignment, enabling robustness across varied room layouts. Quantitative and qualitative evaluations demonstrate improved perceptual quality, structural fidelity, and stylistic consistency compared with conventional template-based visualization methods. The results indicate that diffusion-based generative models provide an effective and scalable solution for automated interior design synthesis with minimal human intervention.

Keywords: Latent Diffusion Models; Image-to-Image Translation; Automated Interior Design; Style-Conditioned Generation; Stable Diffusion.

Abstract: We need to make sure that water filters are safe and work properly. This is important for keeping our water clean and people healthy. If there are leaks or problems with the filters that we do not find they will not work well as they should and people could get sick.This project is, about making a machine to test water filters. The machine is called an automated Pressure Decay Test Bench. It uses a computer system to control everything. It is connected to a program called LabVIEW.The machine tests the water filter by filling it with air and then checking to see how fast the air leaks out over a certain amount of time. This helps us know if the water filter is working correctly. We use the water filter to make sure our water is clean and the machine helps us know if the filter is doing its job. The water filter is an important thing and the machine is helping us to keep it safe. When the pressure drops much we know there is a leak. This way we can find the smallest problems. We use an ABB PM556 PLC to automate things, which means people do not make mistakes. This also means we get the results every time and we can test things faster than if we did it by hand. We use LabVIEW to watch what is happening in time to get the data to see what it means and to store the results of the tests. The system has parts, like pressure transmitters, solenoid valves, manifolds and a compressor. These parts help us control the pressure precisely and measure it accurately. We use the ABB PM556 PLC and LabVIEW to make sure the pressure control and measurement system works correctly. Experimental results demonstrate that the developed test bench provides reliable, efficient, and repeatable leak detection, making it suitable for industrial quality control applications in water filter manufacturing.

Keywords: Pressure Decay Test, Water Filter Testing, Leak Detection, PLC Automation, LabVIEW, Quality Control, Pressure Monitoring.

Abstract: The rapid growth of digital infrastructure and online services has led to a significant increase in the frequency, scale, and sophistication of cyberattacks. Traditional security mechanisms such as firewalls, signature-based intrusion detection systems, and antivirus software are no longer sufficient to detect modern threats like phishing, distributed denial-of-service (DDoS) attacks, brute-force login attempts, and malicious websites. These systems often suffer from high false positives, lack of real-time analysis, and limited contextual understanding. This paper presents CyberShield AI, an AI-powered cyber threat detection and monitoring system designed to provide real-time visibility into multiple cybersecurity threats through an integrated and interactive dashboard. The proposed system analyzes phishing emails, DDoS traffic anomalies, brute-force authentication attempts, and malicious URLs using rule-based intelligence, heuristic analysis, and AI-assisted interpretation. CyberShield AI employs a full-stack architecture with secure authentication, structured data storage, and dynamic visualizations to improve threat awareness and response time. Experimental evaluation demonstrates that the system effectively identifies and categorizes cyber threats while providing clear explanations and actionable insights, making it suitable for academic and practical cybersecurity environments.

Keywords: CyberShield AI, Cyber Threat Detection, Phishing Detection, DDoS Attack Monitoring, Brute-Force Detection, Malicious URL Analysis, Artificial Intelligence, Cybersecurity Dashboard

Abstract: The fundamental building blocks of Computer Programming are Data Structures and Algorithms (DSA). The DSA is a very important part of Software Development, Coding Competitions and Technical Job Interviews. [1] At present, there are several online resources where students can find a wide variety of DSA problems; however, the majority of these platforms do not provide systematic guidance on how to learn and practice DSA. To address this issue, we have created CODE-STEP, an artificial intelligence powered DSA Learning System designed to make learning DSA much easier and more organized than other existing resources, as CODE-STEP presents to students a well-structured and step-by-step approach to solving DSA problems, as opposed to providing random selections. [3] CODE-STEP continually tracks each student's learning progress using performance metrics such as accuracy rates, number of attempts, and time required for completion for each problem they are working on. In order for CODE-STEP to provide a streamlined, interactive, high-performance and user-friendly experience, we use modern technology such as React.js for front-end graphical user interface development; Node.js for back-end processing; MongoDB for data storage; and the Gemini API to provide AI assistance. CODE-STEP creates an encouraging and motivating environment for students to learn DSA without feeling overwhelmed by the amount of material that they need to learn.

Keywords: DSA Learning, Artificial Intelligence, Personalized Learning, AI Chatbot, Smart Recommendation System [5].

Abstract: The increasing reliance on web-based transaction systems has intensified security threats related to anomalous transaction behaviors that may indicate cyberattacks, fraud, or policy violations. Conventional detection mechanisms, such as regex-based filtering and manually defined rule-based methods, are widely used but often suffer from high false positive rates, limited adaptability, and performance instability as transaction patterns evolve. Learning-based approaches offer adaptability but introduce challenges related to explainability, training data dependency, and computational overhead, which limit their suitability for audit-oriented security environments. To address these limitations, this study proposes an automata-based model for transaction anomaly detection integrated with blockchain-based digital evidence storage. The proposed approach models valid web transaction syntax using deterministic finite automata (DFA), enabling transparent, rule-driven anomaly detection without reliance on training data. Blockchain technology is employed as an immutable logging layer to preserve digital evidence of detected anomalies, ensuring integrity, traceability, and auditability. The model is evaluated using simulated HTTP transaction datasets in a local server environment and benchmarked against regex-based and manual rule-based detection methods. Performance evaluation focuses on detection accuracy, false positive rate, execution time, and blockchain overhead in terms of latency and storage consumption. Experimental results demonstrate that the DFA-based model achieves higher detection accuracy, lower false positive rates, and more stable execution times than baseline approaches. Although blockchain integration introduces additional overhead, the impact remains predictable and manageable. Overall, the results indicate that combining automata-based detection with blockchain-based evidence storage provides an effective, explainable, and trustworthy solution for secure web transaction monitoring.

Keywords: anomaly detection; automata-based detection; blockchain evidence storage; transaction security; web transactions.

Abstract: Hematological malignancies, specifically Leukemia, manifest through abnormal white cell proliferation in the bone marrow. Diagnosing this quickly is key for survival. However, looking at slides manually is slow and errors occur. This study works on a dual-stage framework. It couples Particle Swarm Optimization (PSO) with a ResNet-18 back- bone. The architecture handles multi-class classification (ALL, AML, CLL, CML) and severity grading (Grades 1-3) at the same time. PSO functionality is used for hyperparameter tuning. This happens before feature extraction. Validation metrics indicate a precision maximum of 94.2%.

Keywords: Leukemia, Deep Learning, ResNet-18, PSO, Classification, Grading.

Abstract: Disaster scenarios such as earthquakes, landslides, and building collapses create hazardous environments that limit human access and delay rescue operations. Rapid situational assessment and early survivor detection are critical to minimizing casualties. This paper presents the design and implementation of an autonomous hybrid quadcopter–rover system intended for efficient disaster response applications. The proposed system combines the advantages of aerial mobility and ground navigation within a single robotic platform. The quadcopter module enables fast aerial surveillance, real-time video transmission, and terrain assessment, while the rover module facilitates ground-level exploration in confined and debris-filled environments. The system integrates multiple sensors including GPS, IMU, ultrasonic sensors, and cameras for navigation, obstacle detection, and environmental monitoring. Wireless communication allows real-time control and data transmission to rescue teams. Experimental results demonstrate stable aerial performance with payload, reliable rover mobility, low-latency video streaming, and effective hybrid operation. The proposed system reduces human risk, improves accessibility in disaster zones, and enhances the efficiency of search-and-rescue missions.

Keywords: Disaster Management, Hybrid UAV–UGV, Quadcopter Rover, Search and Rescue, Autonomous Robotics, Surveillance System

Abstract: In the era of high-performance computing and miniaturized electronic systems, the demand for high-speed, low- power, and area-efficient arithmetic units has increased significantly. Among all arithmetic operations, multiplication plays a vital role in determining overall system performance, especially in digital signal processing, cryptography, image processing, and embedded systems. Conventional multiplier architectures such as array multipliers, Booth multipliers, and Wallace tree multipliers often suffer from high propagation delay, increased power consumption, and larger silicon area when scaled to higher bit-width operations. To address these limitations, this paper presents the design and VLSI implementation of a 64-bit multiplier based on the Urdhva Tiryakbhayam (UT) Sutra of Vedic Mathematics. The UT algorithm, meaning “Vertically and Crosswise,” enables parallel generation of partial products, resulting in reduced computation delay and improved throughput. The proposed architecture is designed using Verilog HDL and follows a hierarchical approach by decomposing large-bit multiplication into smaller modular blocks. Functional verification is carried out through extensive simulation, followed by synthesis and physical implementation using standard VLSI design flow. Post-synthesis and post-layout analysis demonstrate that the UT-based multiplier achieves superior speed performance while maintaining competitive power consumption and area utilization when compared with conventional multiplier architectures. The results validate the effectiveness of integrating Vedic mathematical principles with modern VLSI methodologies for high- performance arithmetic circuit design.

Keywords: Vedic Mathematics, Urdhva Tiryakbhayam, VLSI Design, High-Speed Multiplier, Verilog HDL

Abstract: This project present an Loan Approval Prediction Using Machine Learning designed to automate and enhance the traditional loan eligibility evaluation process using machine learning and artificial intelligence. Conventional loan approval systems are time consuming, require physical branch visits, and lack transparency in decision-making. To overcome these limitations, the proposed system provides an intelligent, web-based solution that delivers instant loan eligibility predictions without requiring user login. The system evaluates loan applications using a hybrid decision approach, combining rulebased logic and machine learning models to ensure accurate, unbiased, and explainable decisions. Users input basic financial details such as income, loan amount, credit score, and existing liabilities, after which the system calculates EMI, debt-to-income ratio, and generates an approval or rejection result with clear explanations. In addition to loan eligibility prediction, the platform integrates an AI-powered chatbot that assists users with eligibility checks, EMI calculations, branch information, and support queries. A real-time branch locator using map services helps users identify nearby bank branches and obtain navigation directions, improving accessibility and user convenience. The application follows a privacy-first design, ensuring secure handling of data and eliminating unnecessary authentication requirements. The project demonstrates the practical application of machine learning, AI-based automation, and modern web technologies in the banking domain. It provides a scalable foundation for future enhancements such as integration with real banking APIs, document verification, multilingual support, and mobile application deployment, making it suitable for real-world digital banking environments.

Keywords: Loan Approval Prediction, Machine Learning, Artificial Intelligence, Digital Banking, EMI Calculation, Credit Risk Assessment, AI Chatbot, Financial Automation, Banking Assistant, Loan Eligibility System, User-Centric Design, Explainable AI.

Abstract: This paper presents a fully integrated AI-Based Health Monitoring System that performs natural-language symptom analysis, machine learning classification, severity estimation, multilingual recommendation delivery, personal health record management, and real-time emergency support aligned with UN SDG-3. The system accepts free-text symptoms, processes them through TF-IDF vectorization and Logistic Regression, and combines predictive results with rule-based severity logic to ensure medically responsible triage. To improve accessibility, the system provides multilingual outputs in English, Hindi, and Kannada using an offline translation dictionary. Emergency response capabilities include geolocation-based hospital discovery using OpenStreetMap APIs, prioritizing hospitals with verified contact details. Implemented with Python Flask, scikit-learn, SQLite/PostgreSQL, HTML, CSS, and JavaScript, the platform provides safe and reliable triage with intentional over-triage to reduce false negatives in critical conditions.

Keywords: AI Healthcare, Symptom Analysis, Logistic Regression, Multilingual Health System, Emergency Support

Abstract: A mass air flow (MAF) system, based on MD0550 sensor, FPGAs and VHDL, is presented here. The system is capable of providing a series of controls and subsequently activate respective alarm systems, related to car engine operation optimization. It can simultaneously monitor four basic air flow-related parameters. The first is air flow values lowering below lower critical set value, leading to respective insufficient fuel supply. Both blue LED and half left of board LEDs light up and buzzer also sounds to indicate the above fact with simultaneous white LED lighting, which represents additional fuel supply system activation. Second operation parameter checks if air flow values are within set limits and if this holds true, green LED lights up. Third basic operation parameter is related to air flow values becoming higher than upper set values, leading to red LED lighting. Finally, fourth basic operation parameter is checking whether air flow values remain above upper set value for a specific time period, leading to subsequent activation of decrease fuel supply system, represented with lighting yellow LED, while half right of board LEDs also light up. MD0550 air flow sensor used here and its output analog voltage values act as input to FPGA’s ADC unit and converted values are pre-sented to seven-segment displays. The system uses DE10-Lite FPGA board and taking into account that specific time periods, as well as upper and lower air flow limits can be set to a variety of values, gives our system the ability of im-plementation in a wide range of applications such as car engine operation, human breath detection, room occupancy detection, HVAC (Heating, Ventilation and Air Conditioning) system monitoring and weather stations. Our system is cheap to manufacture and can be also combined with IoT systems, allowing its use to logistics applications.

Keywords: MD0550 air flow sensor, air flow monitoring, FPGA, VHDL, Buzzer, LEDs.

Abstract: Charger design for e-bike applications. One big push toward greener travel means more people are riding electric bikes. Plugging them in every time can be awkward, messy, even risky after a while. Cables fray, outlets aren’t always nearby, hands get cold in winter when fiddling with connectors. Instead of cords snaking around wheels and frames, imagine just rolling up close to charge. This idea uses invisible magnetic fields to move energy through thin air between two matching coils. Close but not touching - like magic, only physics. Built into bike and dock alike, these parts pass electricity without sparks or plugs. Less hassle, fewer broken ports, less chance for shocks on wet days. Charging happens quietly, steadily, simply by parking right. A spinning current in the charging pad creates a shifting magnetic push. This energy jumps across space when the bike's pickup ring lines up just right. From there, the captured electricity gets smoothed into usable power for the battery. A small brain made of circuits watches how much juice flows, plus heat levels during the whole process. Too much pressure, too much flow, or rising warmth triggers automatic shields inside the hardware. These layers stop damage before it happens. The setup keeps things steady without needing physical plugs. No cords mean less wear over time, especially where weather or heavy foot traffic could damage equipment. Once the bicycle sits above the pad, power begins without buttons or plugs. Efficiency dips a little because energy jumps across a gap, yet that trade brings fewer breakdowns and simpler daily use. What stands out most is how smoothly it fits into real-world routines. One big win? Cutting cords while juicing up e-bikes out in the open. Power moves through air now, no plugs needed. Safer sidewalks happen when cables vanish. Cities could tuck these pads under bike racks downtown. Think bus stops humming with silent recharge zones. Even rainy days won’t stop current hopping gapless to batteries. Future rides might never need outlets at all.

Keywords: Wireless charging for e-bikes using embedded systems and power electronics.

Abstract: The number of two-wheelers is rapidly increasing. The number of traffic violations and accidents are also becoming substantially higher due to unsafe riding like riding without helmet, triple riding and using mobile phone while riding. The current system for monitoring traffic mainly involves manual monitoring or analog cctvs which affects efficiency and human error, a tedious and time taking the process. To tackle these challenges, the project proposes an AI-Powered Traffic Management System that automatically detects and penalizes traffic violations by two-wheelers, employing the live feed of a webcam or CCTV camera. The YOLOv8 deep learning-based object detection model will be used in the system for quickly detecting helmet absence, triple riding, mobile phone usage while riding, etc. The EasyOCR technique is employed to detect a vehicle’s number plate and then the details of the violation, the details of the rider, date and time, and the fine amount to be charged are all stored in one central location. After the fine is imposed, a system-generated email notification is sent to the vehicle owner containing details of the violation and the amount for which the fine is imposed. All fined records are maintained in the system for monitoring and reporting and for further analysis in the future. The whole system has been developed using Python, OpenCV, Decision Tree and EasyOCR implementing in Visual Studio Code. Moreover, the experimental result shows high accurate, real-time, high accuracy, and reliable detection in practical traffic. The system reduces human interference, automates the traffic rule compliance monitoring, enhances transparency and scalability which boosts road safety and facilitates intelligent and smart traffic management systems (TMS).

Keywords: AI-Powered Traffic Management System, Traffic Violation Detection, YOLOv8, Computer Vision, Helmet and Rider Safety, License Plate Recognition, EasyOCR, Real-Time Video Analysis.

Abstract: Blood donation management systems are critical healthcare applications that ensure the timely availability of safe blood for patients. Most existing systems rely on text-based interfaces, which may limit accessibility during emergencies or for users with visual impairments and low technical literacy. This paper presents an AI-enabled blood donation management system implemented on a localhost platform, integrated with an intelligent voiceover text-to-speech (TTS) module. The proposed system converts system-generated notifications, instructions, and alerts into natural-sounding speech, thereby enhancing accessibility and usability. Unlike cloud-based solutions, the system operates entirely on a local server, making it suitable for small hospitals and blood banks with limited internet dependency. Experimental observations show improved user interaction, faster response to emergency alerts, and reduced operational errors.

Keywords: Blood Donation System, Artificial Intelligence, Text-to-Speech, Multi-Language Voice Assistance, Localhost Platform, Healthcare Applications

Abstract: Road safety remains a critical concern in modern transportation systems, with two-wheeler accidents contributing significantly to traffic fatalities worldwide. This paper presents an intelligent helmet system that integrates multiple safety features including alcohol detection, accident detection, and vehicle ignition control to enhance rider safety. The proposed system employs an MQ-3 alcohol sensor for real-time breath alcohol concentration monitoring, an MPU-6050 accelerometer and gyroscope for accident detection through impact and orientation analysis, and GPS-GSM modules for emergency location tracking and alert transmission. The helmet incorporates an Arduino microcontroller as the central processing unit, coordinating sensor data acquisition, processing, and actuation. The ignition control mechanism prevents vehicle startup when alcohol is detected above legal thresholds, while the accident detection algorithm automatically alerts emergency contacts with precise GPS coordinates upon crash detection. Experimental results demonstrate ninety-two percent accuracy in alcohol detection, ninety-five percent accuracy in accident identification, and response times under two seconds for emergency alert transmission. The system successfully integrates these functionalities in a compact, wearable form factor suitable for practical deployment. This research contributes to intelligent transportation systems by providing a cost-effective, multi-functional safety solution that addresses multiple risk factors simultaneously, potentially reducing accident rates and improving emergency response effectiveness for two-wheeler riders.

Keywords: Intelligent helmet, alcohol detection, accident detection, ignition control, IoT, Arduino, MPU-6050, GPS-GSM, rider safety, embedded systems.

Abstract: WildGuard is an intelligent, real-time surveillance system designed to protect wildlife and forest ecosystems. It integrates state-of-the-art computer vision using YOLOv8 and audio classification based on MFCC features and machine learning classifiers to detect humans, vehicles, animals, and gunshot sounds from live camera and microphone feeds. The system automates threat detection, provides instant email alerts, and stores event evidence in a centralized database with a web-based dashboard for monitoring and analysis. This paper presents the system objectives, architecture, modules, algorithms, implementation details, and experimental outcomes.

Keywords: Wildlife Monitoring, YOLOv8, MFCC, Audio Classification, Real-Time Detection, Flask

Abstract: Water scarcity, unpredictable weather patterns, and inefficient agricultural water management continue to pose significant challenges to global food production. In many farming regions, traditional irrigation methods still depend on manual judgment or fixed scheduling, which often results in excessive watering, uneven moisture distribution, nutrient leaching, and long-term soil degradation. These issues not only waste valuable water resources but also increase operational costs and reduce crop yield. To address these limitations, this project introduces an advanced smart irrigation system that integrates real-time environmental monitoring with a machine learning– driven weather prediction framework to automate and optimize irrigation decisions. A relay-driven water pump mechanism enables automated control of irrigation hardware, eliminating the need for human supervision. When the system detects adequate soil moisture or forecasts expected rainfall, it postpones or stops irrigation to prevent water wastage. Conversely, when data indicates dry conditions or high evapotranspiration rates, the system activates the pump to maintain optimal soil moisture levels for crop growth. This intelligent decision-making significantly reduces water consumption, improves crop health, and enhances overall farm productivity. The prototype results show that combining IoT hardware with predictive analytics creates a highly efficient, scalable, and adaptable irrigation method suitable for both small farms and large agricultural operations. By leveraging machine learning models, the system can continuously improve its prediction accuracy over time, making it a robust solution for climate-resilient agriculture. Ultimately, this smart irrigation framework demonstrates how modern sensing technologies and data-driven automation can transform conventional farming practices into more sustainable, resource-efficient, and environmentally friendly systems.

Keywords: Internet of Things (IoT), Wireless Sensor Networks, Machine Learning, Weather Prediction, Smart Agriculture, Precision Farming

Abstract: Phishing attacks persist as one of the most financially damaging cybersecurity threats, with recent global cybersecurity reports indicating a continuous year-over-year surge driven by large-scale automated phishing kits and AI- generated scam content. Standard reactive defenses like blacklists and static filters simply can’t keep up with modern threats, failing specifically against zero-day attacks that leverage fresh domains or complex social engineering. A workable hybrid approach that uses both URL patterns and email content to identify phishing is needed to close this gap. In order to detect threats without any blacklist entries, the firstlay er employs a logistic regression model – character level TF-IDF vectorization to identify malicious sequence of n-grams 3 to 5 characters.The second layer is an email phishing detection laye r that uses a Random Forest Classifier trained on a UCI Spam base dataset with 57 markers, including word frequencies and capitalization patterns, to identify spam email contents. To avoid false flagging and promptly identify reliable websites, a whitelist is utilized. Both models are managed by the system, which is implemented as a Flask web application. By identifying both phishing URLs and spam patterns, the training results demonstrate the system's high detection rate and low false positives.

Keywords: phishing detection, TF-IDF, logistic regression, random forest, spambase, cybersecurity, URL analysis, and email security

Abstract: Preparing for job interviews effectively remains a critical challenge for job seekers, with many relying on traditional coaching methods that lack personalization, scalability, and real-time feedback mechanisms. Current interview preparation tools provide limited domain-specific guidance, lack interactive real-time evaluation, and fail to capture the nuanced assessment criteria that technical interviewers use. The absence of adaptive, AI-driven mock interview systems leaves candidates underprepared for behavioral, technical, and situational questions tailored to their target roles and experience levels. To address these limitations, the AI-Powered Mock Interview Application integrates Generative AI, Large Language Models (LLMs), and Machine Learning to deliver personalized, interactive interview preparation at scale. The system leverages advanced NLP and transformer-based models to generate contextually relevant technical and behavioral questions based on job role, tech stack, and years of experience provided by users. Real-time speech-to-text conversion captures candidate responses, while AI-powered evaluation mechanisms assess communication clarity, technical accuracy, and confidence levels against industry benchmarks. The application provides instant, detailed feedback including answer quality assessment, improvement suggestions, performance analytics, and comparative metrics across multiple interview attempts. Through a user-centric web platform, candidates access role-specific question banks, view personalized performance dashboards, and receive AI-generated recommendations for skill enhancement. Stakeholders including job seekers, career coaches, and educational institutions benefit from comprehensive analytics and interview metrics. By combining adaptive question generation with real-time speech analysis and predictive performance insights, the proposed solution significantly improves candidate confidence, reduces interview anxiety, and enhances job selection probability while democratizing access to high-quality interview preparation.

Abstract: Managing biomedical waste safely is one of the toughest challenges faced by healthcare facilities, especially because manual segregation exposes workers to significant risks. To address this, our project introduces an automated medical waste sorting system designed to reduce human involvement and improve safety. At the heart of the system is a YOLO-based object detection model, which can identify commonly discarded medical items such as syringes, gloves, cotton pads, and masks using live camera input. Once an item is recognized, a Rasp berry Pi–powered robotic arm takes over, performing contactless pick-and-place operations to sort the waste into the correct bins. We tested the system under realistic operating conditions, and it consistently delivered accurate detection along with reliable robotic performance. These results demonstrate how combining deep learning with robotics can create a safer, more efficient approach to biomedical waste management, paving the way for smarter healthcare practices in the future.

Keywords: Medical Waste Management, Automated Waste Segregation, Object Detection, YOLO, Robotic Arm Automation, Raspberry Pi, Deep Learning

Abstract: Urban traffic management has become one of the most critical challenges facing modern cities as rapid urbanization and exponential growth in vehicle ownership continue to strain existing transportation infrastructure. Traffic rule violations significantly impact road safety and public welfare, and traditional manual monitoring methods are often slow, inconsistent, and dependent on human personnel who suffer from fatigue and limited visibility. To address these challenges, the proposed system leverages deep learning to detect and classify violations from video footage with improved precision and reliability. A comprehensive approach using YOLOv8 for object detection and EasyOCR for license plate recognition was implemented and validated across diverse traffic conditions. The model effectively extracts spatial and temporal features from input video frames and achieves high performance, recording approximately 96.8% vehicle detection accuracy and 95.3% overall violation classification accuracy. The solution is deployed as an interactive web application built with FastAPI, enabling traffic authorities—particularly enforcement officers and urban planners—to upload footage and receive real-time violation alerts. By offering a fast, affordable, and scalable enforcement tool, this work contributes to smarter traffic management practices, timely violation detection, reduced dependency on manual monitoring, and overall enhancement of urban road safety. The study also highlights the potential of YOLOv8-based systems to transform traditional traffic law enforcement through efficient, user-friendly, and technology-driven approaches.

Abstract: The recent boom in the electric vehicle (EV) industry has escalated the demand towards effective and safe electric vehicle charging solutions that are also convenient to use. The conventional wired charging systems are based on physical connectors that are vulnerable to mechanical damages, are unsafe and may be inconvenient to the drivers. We propose a solution to these shortcomings by introducing a static wireless charging system, which incorporates the Internet-of-Things (IoT) feature to monitor and control it in real-time. In our system, the inductive power transfer is used: an AC source with high-frequency is produced by using a power-electronics converter on the mains, which is fed to a transmitter coil. The alternating magnetic field is coupled to a receiver coil on the EV to induce an AC voltage which is then rectified, filtered and regulated to offer a constant DC supply to charge the batteries. An embedded Wi-Fi/BLE chip and microcontroller allow nonstop monitoring of such vital parameters as charging voltage, current, temperature, and status to a cloud-based dashboard, which contributes to a higher level of safety, diagnostics, and user experience. A prototype was also made and tested under stationary conditions. On-hand experimental findings provide credible transfer of power over a small air gap, with the preservation of output voltage and successful battery charging. The concept of real-time monitoring of IoT was found to be effective in delivering real-time data about the performance of the charging process and allowing faults to be identified remotely. The proposed design will save wear and tear on connectors, eradicate cable risks and increase the general convenience of the system in comparison to the standard wired system.The results of this research allow recognizing the possibility of combining the technique of charging an electric vehicle in place with the IoT-based monitoring system and positioning it as an attractive element of the following smart and automated charging systems.

Keywords: IoT, Electric vehicles, Wireless charging, Static charging, Wireless Power Transmission (WPT), Magnetic field

Abstract: This paper presents the design and implementation of an AI-Powered Career Intelligence Platform aimed at enhancing employability through intelligent interviewing, adaptive coding practice, automated resume optimization, and collaborative learning. The proposed system integrates Artificial Intelligence (AI), Natural Language Processing (NLP), and Machine Learning (ML) techniques to deliver personalized, data-driven assistance for students and professionals seeking to improve career readiness. The platform is structured into four core modules. The first module, the AI-Based Interview System , automatically generates role-specific questions, records user responses through speech, converts them to text using multilingual speech recognition, and provides real-time, performance-based feedback using sentiment and semantic analysis. The second module, the Adaptive DSA Learning Engine, delivers context-aware coding exercises with dynamic difficulty adjustment, interactive hints, and AI-assisted code evaluation for personalized skill development. The third module, the ATS-Driven Resume Builder, optimizes resumes using real-time scoring, tone adjustment, and keyword enrichment aligned with Applicant Tracking System (ATS) standards. The fourth module, the AI-Enhanced Community Platform, employs knowledge graphs, peer endorsements, and automated discussion summarization to foster collaborative learning and professional networking. With multilingual support for Kannada, Hindi, and English, the system ensures inclusivity across diverse user groups. By combining adaptive analytics with NLP-driven insights, the platform establishes a unified, scalable, and intelligent ecosystem that bridges education, career preparation, and continuous employability enhancement.

Keywords: Artificial Intelligence (AI), Natural Language Processing (NLP), Machine Learning (ML), Adaptive Learning, Employability Enhancement, Resume Optimization, Knowledge Graphs.

Abstract: This paper presents CORTEX, an automated security analysis and response framework designed to mitigate analyst burnout and streamline the investigation of security observables. By centralizing disparate intelligence sources into a unified command center, the system eliminates "tab-switching fatigue" through the concurrent execution of specialized "Analyzers" across global security databases. CORTEX transforms raw data into actionable intelligence in seconds, enabling defenders to pivot from manual data collection to high-level strategic decision-making. Furthermore, the integration of an active "Responders" module facilitates a seamless transition from passive detection to automated mitigation, such as system isolation and user blocking. Ultimately, this framework optimizes the incident response lifecycle, providing security teams with a scalable, human-centric workflow to counter sophisticated cyber threats.

Keywords: Security Automation, Observable Analysis, Incident Response, Threat Intelligence, Operational Efficiency, CORTEX Framework.

Abstract: Modern software systems need to be scalable, deploy quickly, and offer strong visibility. Traditional monolithic applications struggle to meet these needs because they are tightly connected and complicated to deploy. This work introduces a cloud-native microservices-based e-commerce platform made with Docker, Kubernetes, and DevOps CI/CD automation. It has full visibility through Open Telemetry, Prometheus, and Grafana. Each business service—Product Catalog, Cart, Checkout, Payment, Shipping, Email, and Recommendation—is independent and deployed separately. Kubernetes handles load balancing, scaling, and self-repair. GitHub Actions automates the CI/CD pipeline for testing, building, and deploying. Visibility tools offer distributed tracing, metrics, and logs for debugging and checking performance. Experimental results show faster deployment speeds, lower latency, and better traceability across services.

Abstract: Artificial Intelligence (AI) is redefining how public sector banks operate in India by introducing smart Decision Support Systems (DSS) that enhance efficiency and accuracy in decision-making. This research paper investigates how AI-based DSS contributes to operational efficiency in public sector banks, especially in core banking operations, risk management, customer service, and process automation. The study context is grounded in the Indian banking landscape, where digital transformation has accelerated due to competitive pressures and customer expectations. The research synthesizes primary concepts of AI, DSS frameworks, and operational efficiency metrics to understand the depth of AI adoption and its practical effects on public sector banks. Review of recent Indian academic research shows significant positive correlations between AI integration and improvements in service delivery, workflows, risk mitigation, and back-office task optimization. However, implementation challenges such as data quality, regulatory constraints, technical know-how, and ethical considerations persist. By employing a mixed-method methodology combining descriptive analysis and inferential evidence from secondary sources, this paper discusses both qualitative and quantitative implications of AI-driven DSS. It concludes that while AI-enabled systems can considerably reduce processing time, errors, and operational costs, the banks need supportive infrastructure, skilled workforce training, and robust governance frameworks to fully leverage these systems. The study’s findings provide actionable insights for policymakers, banking executives, and technology strategists to enhance operational efficiency and guide future research on advanced AI integration in public sector banking.

Keywords: Artificial Intelligence, Decision Support Systems, Public Sector Banks, Operational Efficiency, India

Abstract: Urban communities routinely discard reusable goods due to limited access to structured channels for redistribution. Many individuals and families, meanwhile, experience a shortage of basic resources that could be filled using existing surplus. This work presents a mobile application designed to enable peer-to-peer donation and exchange of household items, educational materials, wearable goods, and electronic equipment. The system supports listing, request handling, status tracking, and real-time messaging between participants, thereby transforming informal donation practices into a transparent and traceable process. Developed as a native Android solution using Kotlin, Firebase, and Retrofit APIs, the application was evaluated in terms of performance, usability, and user perception. Results indicate that the system supports efficient onboarding, rapid synchronization of data, and smooth communication with minimal user effort, making it well-suited for community-oriented digital welfare initiatives.

Keywords: Donation system, Peer-to-peer exchange, Android development, Firebase, Sustainability.

Abstract: Communication barriers between hearing-impaired individuals and the general population pose significant challenges in education, healthcare, and daily interactions. Sign language serves as an essential medium for such communities, yet the lack of widespread proficiency creates a persistent accessibility gap. Recent progress in computer vision and deep learning provides a promising pathway to automate the interpretation of sign gestures in real time. This work presents a Convolutional Neural Network (CNN)-based sign language recognition framework that accurately classifies static hand gestures from the American Sign Language (ASL) alphabet. The system integrates image preprocessing, region-of-interest (ROI) extraction, and optimized feature learning to enhance recognition efficiency under varying lighting, backgrounds, and hand orientations. To build a robust model, multiple CNN architectures—including MobileNetV2, a custom deep CNN, and a classical LeNet-5 variant—were trained and evaluated on the ASL Alphabet Dataset. An ensemble fusion mechanism was designed to combine the predictive strengths of all three networks, producing a stable and highly accurate classification output. Post-processing with an N-gram–based decoder further improves consistency by reducing misclassification of visually similar signs. Experimental evaluation demonstrates that the proposed approach delivers strong performance across key metrics such as accuracy, precision, recall, and inference time, enabling reliable real-time deployment. The resulting system supports text-based and text-to-speech outputs, offering a practical tool for inclusive communication. Overall, the research provides a scalable and efficient solution for sign language recognition, contributing toward accessible human–computer interaction technologies.

Abstract: Early detection of plant diseases is crucial for reducing economic losses and ensuring global food security. Traditional visual inspections by farmers are subjective and time-consuming, prompting the need for automated solutions. This paper presents a machine learning-based system for identifying and classifying plant leaf diseases using convolutional neural networks (CNNs). We describe the preprocessing, augmentation, and segmentation techniques employed to enhance data quality and improve model performance. Our experiments, conducted on a dataset of 90,000 images across 38 classes, achieved a training accuracy and a validation accuracy above 98% The system also features an intuitive web interface for practical deployment, supporting real-time detection in agricultural fields.

Keywords: Plant Disease Detection, Agriculture, Machine Learning, Deep Learning, Convolutional Neural Networks (CNN), Image Processing, Data Augmentation, Image Segmentation, Ensemble Learning, VGG16, VGG19, ResNet101V2, InceptionV3, LIME Explainability, Computer Vision, Transfer Learning, Attention Mechanisms, Public Datasets, Real-Time Detection, Web-Based System, Smart Agriculture, Food Security.

Abstract: In today’s increasingly interconnected digital environment, protecting information systems from evolving cyber threats has become a critical necessity. Traditional intrusion detection and prevention systems often rely on static, rule-based approaches, which are insufficient to identify sophisticated and previously unseen attacks. This paper presents an Intelligent Intrusion Prevention System (IPS) that leverages machine learning techniques for advanced data protection and real-time threat analysis. The proposed system is deployed within a self-hosted Linux-based private cloud environment, created by converting a standard laptop into a secure server infrastructure, ensuring data sovereignty and administrative control. Machine learning models such as Random Forest, Support Vector Machine (SVM), and Deep Neural Networks (DNN) are employed to analyse network traffic and system logs, enabling accurate classification and prediction of intrusion activities. To enhance security, the system integrates encryption mechanisms, access control policies, and automated firewall actions, providing a multi-layered defence framework. A real-time monitoring dashboard visualizes intrusion attempts, system performance, and threat metrics, allowing prompt response and mitigation. The results demonstrate that the proposed system delivers an adaptive, scalable, and cost-effective cybersecurity solution capable of autonomous threat detection and prevention while preserving data privacy. This approach offers a practical and efficient framework suitable for academic, research, and small organizational environments.

Keywords: IPS, DNN, Real-Time Threat Analysis, Cybersecurity, Anomaly Detection.

Abstract: GlobalTalk is a multilingual communication system designed to make interaction across languages smoother and more accessible. It brings together text-to-speech, translation, and pitch-controlled speech output in one unified platform. Built using Flask for backend routing and the Web Speech Synthesis API for real-time voice generation, the system allows users to log in, choose from available system voices, adjust pitch levels, and listen to text in clear, natural-sounding speech. Its translation feature, powered by Deep Translator, supports automatic language detection and enables easy conversion between multiple languages. The interface is simple and responsive, thanks to clean HTML/CSS layouts and modular JavaScript functions. User activity, such as spoken text and selected voices, is stored in memory to deliver a smooth and personalized experience. Overall, GlobalTalk offers an efficient and user-friendly environment for multilingual communication, while also providing a strong base for future enhancements like neural TTS, multi-agent extensions, and real-time conversational translation.

Abstract: This paper introduces Prompt2Extension, that allows users to create functional, installable browser extensions using natural language prompting. Even though the power of the large language models (LLMs) has shown a tremendous effect on the software development process, designing user-specific browser extensions in an automated manner is not an easy task. The sophisticated, prompt-centric architecture of our system can convert high-level user requirements into multi file browser extensions with massive scale and using the Gemini 1.5 Pro API of Google. We rely on a complex Chain-of-Thought system prompt that forces the LLM to strategize the required files and permissions, judge the request and explain the logic prior to creating the final code. The formatted JSON object created by this one-shot generation process contains all required files (manifest.json, JavaScript, HTML and CSS) and is bundled into a downloadable .zip file. One of the features of our React-based user interface is the ability of the user to look at the generated code before installation. Experiments have shown that this prompt-driven method can effectively create a broad spectrum of extensions, including simple style manipulators as well as more complex, event-driven extensions that need background scripts. This piece shows how advanced prompt engineering can become in order to turn a specialized field of software development into a far more accessible field by allowing non-programmers to build their own original web tools.

Keywords: Prompt Engineering, Code Generation, Browser Extensions, Natural Language Processing, Gemini, Large Language Models, Fast-API, React, Chain-of-Thought

Abstract: The increasing demand for automated retail solutions has led to the development of smart checkout systems that eliminate the need for manual billing. Traditional checkout processes are time-consuming and prone to human error, creating a need for automated and reliable systems. This paper presents an object detection-based auto checkout system using the YOLO (You Only Look Once) deep learning model to identify and classify items in real time. The system integrates image acquisition, preprocessing, and YOLO-based detection to provide accurate and fast billing. Experiments conducted on a dataset of over 10,000 images of grocery items demonstrated a detection accuracy above 95%, reducing checkout time and improving customer convenience.

Keywords: Auto Checkout System, YOLO, Object Detection, Computer Vision, Deep Learning, Automated Billing, Smart Retail, Real-Time Detection

Abstract: Recent research work on air quality checker technology highlight significant advancements driven by Internet of Things (IoT) and Artificial Intelligence (AI), the proliferation of low-cost sensors, and the exploration of hybrid monitoring approaches. Air pollution is a most unavoidable serious issue which risks to human health and ecosystems. Continuous monitoring of air quality is essential for assessing pollution levels and for taking timely preventive measures. The standard indicator use for air quality measurement is Air Quality Index (AQI). It includes concentration of major pollutants such as particulate matter (PM₂.₅ and PM₁₀), sulfur dioxide (SO₂), nitrogen dioxide (NO₂), carbon monoxide (CO), and ozone (O₃). This paper reviews various air quality monitoring approaches, including conventional monitoring stations, Internet of Things (IoT)-based systems, and data-driven techniques. Traditional monitoring methods provide high accuracy but suffer from high cost and limited coverage. Recent advancements in low-cost sensors, wireless communication, and cloud platforms have enabled real-time and scalable AQI monitoring. Furthermore, machine learning techniques have improved AQI prediction and pollution trend analysis. The review highlights existing challenges such as sensor calibration, data reliability, and environmental interference, and discusses future research directions for smart and sustainable air quality monitoring systems.

Keywords: Air Quality Index (AQI), Air Pollution Monitoring, IoT, Low-Cost Sensors, Machine Learning, Environmental Monitoring, Smart Cities

Abstract: In this paper, rough hesitant neutrosophic sets are introduced. Also applying this set to multi criteria decision making problem. In addition an algorithm to handle decision making problem in online teaching company to select staff’s are studied. Finally, a numerical example is employed to demonstrate the validness of the proposed rough hesitant neutrosophic sets.

Abstract: Employee attrition poses a significant financial and operational challenge to modern organizations, leading to increased recruitment costs and loss of institutional knowledge. This research proposes a robust predictive framework to identify at-risk employees and determine the underlying drivers of turnover. Utilizing the IBM HR Analytics dataset, we implement a machine learning pipeline centered on the Random Forest Classifier. To address the inherent class imbalance in attrition data, the Synthetic Minority Over-sampling Technique (SMOTE) was employed, significantly improving the model's sensitivity to minority class instances. Experimental results demonstrate that the model achieves an F1-score of [Insert Score, e.g., 0.89] and an AUC-ROC of [Insert Score, e.g., 0.92]. Feature importance analysis identifies Monthly Income, Overtime, and Age as the primary predictors of turnover. The study concludes with the deployment of a web-based dashboard, providing HR practitioners with an actionable tool for proactive intervention and data-driven retention strategies.

Keywords: Machine Learning, Employee Attrition, Random Forest, SMOTE, Predictive Analytics, HR Management.

Abstract: The rapid growth of network-based services has increased the exposure of modern communication infrastructures to a wide range of cyber attacks, making accurate and timely intrusion detection a critical requirement. Traditional rule-based security mechanisms often struggle to detect evolving and multi-type network attacks due to their reliance on predefined signatures. This paper presents a machine learning–based framework for the prediction and classification of multi-type network attacks using time-based traffic features. The proposed system analyzes temporal characteristics of network flows, including packet inter-arrival times, flow duration, and active–idle behavior, to distinguish benign traffic from malicious activities and further classify attacks into specific categories. A trained machine learning model processes network traffic data provided in CSV format and performs multi-class attack classification with high accuracy. An interactive dashboard developed using Python Dash enables users to upload traffic data, execute predictions, and visualize results through charts and detailed tables. Experimental evaluation demonstrates that time-based feature analysis significantly enhances detection performance compared to conventional approaches, while providing an automated, scalable, and user-friendly solution for network security monitoring.

Keywords: Network Attack Detection, Time-Based Traffic Features, Machine Learning, Multi-Class Classification, Intrusion Detection System, Network Security Dashboard

Abstract: The construction industry plays a vital role in economic development and heavily depends on skilled and unskilled labour. However, the process of hiring and managing labour is still largely manual and unorganized. Contractors usually rely on personal contacts, local labour markets, or intermediaries to find workers, while labourers depend on daily availability at fixed locations or through informal networks. This traditional approach often results in delays, inefficiency, lack of transparency, and uncertainty for both parties. Solving this problem is important to improve efficiency, trust, and productivity in the construction sector. A digital system can reduce manual effort, save time, and provide better coordination between contractors and labourers. By introducing features such as profile management, skill-based search, rating systems, and location-based access, both parties can make informed decisions and work more effectively.

Keywords: Labour Management System, Contractor Management, Workforce Management, Web Application, Database System, Smart Hiring.

Abstract: The rise in fake diplomas causes big problems for schools, companies, and groups that check credentials. Because old-school checks rely on central record systems and hand reviews, they tend to be sluggish, costly, while opening doors to tampering. Instead of sticking with those outdated methods, this study introduces SecureCert - a system that issues and confirms certificates across a decentralized network using Ethereum’s blockchain, file hosting via IPFS, data protection through SHA-256 encryption, identity checks with scannable QR tags, along with an automated tool made in Python. The system lets schools make PDF certificates, then calculate digital fingerprints - after that, they’re sent to IPFS while info gets locked into Ethereum via smart contracts. To check validity, users scan a QR code or type in an ID, which pulls data from the chain and confirms the IPFS hash along the way.

Keywords: Blockchain, Ethereum, SecureCert, Academic Certificate Verification, Decentralized Systems, IPFS, Smart Contracts, SHA-256 Encryption, QR Code Verification, Tamper Proof Records, Digital Credentials

Abstract: This paper presents a machine learning–based Sleep Disorder Prediction System designed to assist in the early identification and awareness of common sleep-related disorders. The system analyzes user-provided health and lifestyle parameters such as sleep duration, body mass index (BMI), stress level, physical activity, heart rate, and blood pressure to predict sleep conditions including Healthy, Insomnia, and Sleep Apnea. By integrating a trained classification model with a Django-based web application, the system provides real-time predictions along with confidence scores to help users better understand their sleep health status. The platform also includes features such as user authentication, prediction history tracking, doctor appointment booking, and contact support, making it practical for real-world usage. This approach demonstrates how data-driven machine learning techniques can offer an accessible, cost-effective, and user-friendly solution for preliminary sleep disorder assessment and promote proactive sleep health management.

Keywords: Sleep Disorder Prediction, Machine Learning, Insomnia, Sleep Apnea, Healthcare Analytics, Web Application.

Abstract: Ensuring women’s safety in urban and semi-urban environments has become a critical societal and technological challenge. Conventional safety mechanisms often rely on manual reporting, delayed emergency calls, or single-channel alert systems, which may fail during high-stress or critical situations. This work presents an AI-Based Women Safety and Alert System that integrates real-time location tracking, intelligent alert processing, and multi-channel communication to provide immediate assistance during emergencies. The proposed system enables users to trigger SOS alerts through a web or mobile interface, automatically sharing live location details with emergency contacts and nearby authorities. Artificial intelligence modules are incorporated to analyze incident patterns and support future prediction of unsafe zones. The system utilizes SMS, Email, and Push Notifications to ensure reliable alert delivery even under network constraints. Administrative dashboards provide real-time monitoring, alert logs, and response tracking for accountability. Experimental evaluation demonstrates that the system significantly reduces emergency response time while improving reliability and transparency compared to traditional safety applications. The results indicate that AI-driven, multi-channel alert systems can play a vital role in proactive women safety solutions.

Keywords: Women Safety, SOS Alert System, Artificial Intelligence, Real-Time Location Tracking, Emergency Response, Smart Safety Systems.

Abstract: The rapid growth of digital reading has increased the demand for web-based platforms that provide a seamless and personalized reading experience. However, existing e-book systems often offer fragmented services where book discovery, reading, progress tracking, and personalization are handled separately. This paper presents a web-based digital reading platform designed to integrate book browsing, PDF reading, bookmarking, highlighting, and automatic progress tracking into a single unified system. The platform is developed using a modern technology stack consisting of React for the frontend, Flask for backend services, and MySQL for persistent data storage, secured with JWT-based authentication. The system enables readers to resume reading from their last saved position, manage bookmarks and highlights, and access external book references through Goodreads and online marketplaces. Experimental testing demonstrates that the platform provides reliable performance, secure access, and an improved user reading experience. The proposed system offers a scalable foundation for future enhancements such as analytics-driven recommendations and advanced annotation tools.

Keywords: E-book platform, Digital reading, PDF viewer, Bookmarking, Reading progress tracking, Web application, Goodreads

Abstract: The Disease Surveillance and Analytics Platform is designed to enhance the monitoring and management of disease outbreaks by providing a centralized and intelligent data processing system. The platform enables healthcare institutions to securely submit disease case information through structured digital channels, ensuring accurate and timely data collection. Once collected, the data is validated, stored, and analyzed to identify trends, patterns, and potential risk scenarios. Interactive dashboards and real-time alert mechanisms facilitate quick interpretation of health data and enable authorities to respond promptly to emerging threats. Secure authentication and role-based access control ensure authorized system usage and data protection. By leveraging modern web technologies and real-time communication frameworks, the platform delivers scalable, reliable, and responsive disease monitoring capabilities. This system demonstrates how data-driven solutions can strengthen public health surveillance and support effective decision-making in real-world healthcare environment.

Keywords: Disease Surveillance, Healthcare Analytics, Real-Time Monitoring, Outbreak Detection, Data Visualization, Public Health Management.

Abstract: The rapid advancement of information technology has led to the integration of intelligent systems into various domains, especially healthcare. With the growing population and increasing number of patients, the demand for quick, accurate, and affordable medical diagnosis has become more critical than ever. Traditional medical diagnosis methods rely heavily on human expertise, which may be time-consuming and prone to errors due to fatigue and workload. To overcome these limitations, the implementation of machine learning and artificial intelligence in medical diagnosis has gained significant attention in recent years.[1] The proposed project focuses on building a software-based intelligent disease prediction system that helps users identify possible diseases based on their symptoms. The system uses machine learning algorithms to analyze symptom input and generate accurate predictions. This approach provides users with fast preliminary health analysis, allowing them to take timely medical action. In rural and remote areas where access to healthcare professionals is limited, such systems play a crucial role in providing early medical guidance.

Abstract: Managing incoming calls during busy or unavailable periods remains a significant challenge in modern organizational environments, often leading to missed important calls and inefficient follow-up scheduling. This paper presents AVI – an AI Virtual Assistant that automates call handling and meeting scheduling by integrating Twilio telephony services, a Fast API backend, Python logic, Google Calendar API, and a React-based dashboard. Incoming calls are processed through Twilio webhooks, where digit-based user inputs are forwarded to the backend for deterministic decision-making and real-time calendar availability checks. Scheduling requests are displayed on the dashboard with caller details and approve or reject options, enabling user-controlled meeting management. Upon approval, a meeting link is automatically generated and updated on the dashboard, while rejected requests are declined without further notification. Experimental results indicate that AVI offers reliable request handling, smooth backend-to-dashboard synchronization, and efficient automated scheduling, making it a scalable and practical solution for intelligent call automation.

Keywords: AI Virtual Assistant, Call Automation, Twilio Webhooks, Fast API Backend, Automated Meeting Scheduling, Google Calendar Integration, React Dashboard, Telephony Automation.

Abstract: Extracting meaningful information from large PDF documents remains a significant challenge for students, researchers, and professionals, as traditional document reading and keyword-based search methods are time-consuming and inefficient. Existing tools often lack semantic understanding, contextual awareness, and interactive capabilities, making it difficult for users to obtain precise answers from complex documents. This limitation results in reduced productivity and increased effort when analysing lengthy technical, academic, or legal PDFs. To overcome these challenges, the AI Powered PDF Chat Application integrates Generative AI, Large Language Models (LLMs), and vector-based semantic search techniques to enable intelligent document interaction. The system processes uploaded PDF documents by extracting text, generating embeddings, and storing them in a vector database to support context-aware retrieval. When a user submits a query, the application identifies the most relevant document segments using similarity search and generates accurate, context-driven responses through an AI language model. This approach ensures that answers are grounded in the document content rather than relying on generic responses. The application is implemented as a secure web-based platform featuring user authentication, PDF preview, interactive chat interface, conversation history management, and PDF-based chat export functionality. By combining retrieval-augmented generation with real-time user interaction, the system significantly improves information accessibility, reduces document analysis time, and enhances user comprehension. The proposed solution demonstrates how AI-driven document intelligence can transform traditional PDF reading into an efficient, interactive, and scalable knowledge retrieval experience.

Abstract: The rapid growth of urbanization and the increasing number of two-wheelers on roads have significantly intensified traffic congestion and safety challenges. Traditional traffic monitoring systems rely heavily on manual observation or basic image processing techniques, which often fail to provide accurate, real-time analysis under complex road conditions. To overcome these limitations, this project presents a computer vision–based traffic object detection system using YOLOv8, designed specifically for bike-only traffic scenarios. The proposed system focuses on detecting critical traffic-related objects, including person, helmet, and vehicle number plate, from surveillance video streams. Video frames are extracted using OpenCV and manually annotated using LabelMe. Since YOLOv8 does not support JSON annotations directly, the annotations are converted into YOLO format with normalized bounding box coordinates. A pretrained YOLOv8 model is fine-tuned on the custom dataset to achieve accurate real-time detection. During inference, the trained model processes video frames and outputs bounding boxes with class labels and confidence scores. Experimental results demonstrate reliable detection performance under both daytime and nighttime conditions, with minimal false detections. The modular architecture of the system enables easy extension for higher-level traffic analysis such as helmet violation detection and number plate recognition.The proposed approach provides an efficient, scalable, and intelligent solution for automated traffic surveillance and serves as a strong foundation for smart transportation systems

Keywords: Traffic Object Detection, YOLOv8, Computer Vision, Helmet Detection, Number Plate Detection, Intelligent Transportation System

Abstract: The rapid growth of intelligent camera applications has increased the demand for automated and user-friendly image capturing systems. Conventional selfie capturing methods require manual interaction, which may lead to poor timing, blurred images, or unnatural facial expressions. To overcome these limitations, this project presents an Auto Selfie Capture by Smile system that automatically captures a photograph when a user smiles. The proposed system utilizes computer vision and machine learning techniques to detect human faces and recognize smiling expressions in real time using a live camera feed. Haar Cascade classifiers are employed for efficient face and smile detection. Once a smile is detected and maintained for a predefined duration, the system automatically captures and stores the selfie image without any user intervention. Additional checks such as face alignment and stability help improve image quality and reduce false captures. This system provides a hands-free, accurate, and efficient solution for selfie capturing, making it suitable for applications in smart cameras, mobile devices, and human–computer interaction systems. The proposed approach is lightweight, cost-effective, and capable of real-time performance on standard hardware.

Keywords: Auto Selfie Capture, Smile Detection, Face Detection, Computer Vision, OpenCV, Haar Cascade Classifier, Image Processing, Human–Computer Interaction.

Abstract: The digital synthesis of human kinematics and gestural linguistics represents a sophisticated frontier in computational intelligence, with profound implications for assistive communication, healthcare diagnostics, and touchless human-computer interaction (HCI). Traditional methodologies for movement analysis frequently encounter a "performance-efficiency" bottleneck, where high-fidelity recognition often requires excessive computational overhead, rendering real-time deployment on standard consumer hardware impractical. Furthermore, conventional pixel-based processing is often compromised by environmental noise, varying illumination, and complex background occlusions. This research introduces an Integrated Multi-Modal Perception Framework that unifies skeletal tracking, behavioral classification, and sign language interpretation into a singular, high-performance ecosystem. The system bypasses the limitations of traditional Convolutional Neural Networks (CNNs) by adopting a landmark-centric approach. By utilizing the MediaPipe perception pipeline, the framework achieves high-fidelity extraction of 33 body landmarks and 21 per-hand keypoints in a 3D coordinate space. This transformation of raw video into a low-dimensional kinematic topology allows for fluid execution without the necessity for dedicated GPU acceleration. To resolve the challenge of interpreting dynamic motion, the system implements a Long Short-Term Memory (LSTM) Recurrent Neural Network. This architecture is specifically engineered to model spatiotemporal dependencies across sequential frames, enabling the system to distinguish between similar but chronologically distinct actions. A defining innovation of this project is its decoupled modular architecture, which facilitates the independent execution of pose and hand modules through a shared, optimized preprocessing stream. The integration of confidence-based thresholding and temporal smoothing further ensures the stability of predictions during live interaction. Empirical testing confirms that the proposed system delivers a robust, low-latency solution capable of operating at real-time frame rates on standard CPU architectures. By democratizing access to advanced gesture recognition, this work contributes to the development of inclusive technology that bridges the gap between physical human movement and digital understanding.

Abstract: Local and grassroots cricket tournaments continue to rely on manual player selection methods such as verbal bidding, paper records, and spreadsheet-based tracking. These traditional approaches lack transparency, are prone to calculation errors, and often result in disputes, biased selections, and inefficient budget management. The absence of real-time validation and post-auction analytics further reduces the credibility and effectiveness of tournament organization. To address these challenges, CrickBidder is proposed as a web-based online cricket player auction platform that digitizes and automates the entire auction workflow. The system enables organizers to conduct IPL-style live auctions where teams can bid for registered players in real time under predefined rules. CrickBidder incorporates role-based access for organizers, teams, and players, ensuring secure participation and controlled auction management. Real-time communication mechanisms ensure instant bid updates, timer synchronization, and transparent winner determination. The platform is developed using modern web technologies, including React.js for the frontend, Node.js with Express.js for backend services, MongoDB for data storage, and Socket.io for real-time bidding updates. Automated bid validation, purse management, and squad constraints reduce human errors and ensure fairness. Post-auction reports and analytics provide valuable insights into team spending and player demand. By digitizing the auction process, CrickBidder enhances transparency, efficiency, and professionalism in grassroots cricket management. The system demonstrates how real-time web technologies can be effectively applied to modernize sports management and deliver a scalable, data-driven auction solution.

Abstract: Agriculture remains the backbone of many developing economies, yet farmers continue to face significant challenges in selecting the most suitable crop for cultivation due to unpredictable climatic conditions, soil variability, and limited access to data-driven decision support systems. Traditional crop advisory methods largely depend on manual expertise, historical practices, or generalized recommendations, which often fail to account for real-time environmental parameters and regional diversity. With the increasing availability of agricultural datasets and advancements in Machine Learning (ML), there is a growing opportunity to enhance crop selection accuracy and improve farming outcomes through intelligent systems. This research presents CropSense_AI, a machine learning–based crop recommendation system designed to assist farmers in identifying the most appropriate crop to cultivate based on key soil and environmental parameters. The system utilizes essential inputs such as nitrogen, phosphorus, potassium levels, soil pH, temperature, humidity, and rainfall to generate reliable crop recommendations. A Random Forest classification algorithm is employed due to its robustness, ability to handle non-linear relationships, and resistance to overfitting when working with real-world agricultural data. The model is trained and evaluated using a well-structured agricultural dataset, achieving high prediction accuracy and stable performance across multiple crop classes. Unlike many existing solutions that rely heavily on IoT sensors, image processing, or complex infrastructure, CropSense_AI focuses on simplicity, accessibility, and interpretability. The system is designed to operate using readily available data inputs, making it suitable for deployment in resource-constrained rural environments. Additionally, the web-based interface allows users to interact easily with the system, visualize input parameters, and understand prediction outcomes without requiring technical expertise. This practical design ensures that the system can be adopted by farmers, agricultural officers, and extension services with minimal training. The proposed system bridges a critical gap in current agricultural decision support tools by combining accuracy, usability, and deployment readiness. By providing data-driven crop recommendations before cultivation, CropSense_AI has the potential to reduce crop failure risk, optimize resource utilization, and support sustainable farming practices. The results demonstrate that machine learning–based crop recommendation systems can play a vital role in modern precision agriculture, contributing to improved productivity, informed decision-making, and long-term agricultural sustainability.

Abstract: Long-term financial planning and property acquisition decisions are often made using fragmented tools such as manual budgeting methods, basic online calculators, or informal financial advice. These approaches fail to provide integrated, personalized, and explainable insights that account for income patterns, expenses, liabilities, and future financial commitments. This paper presents Legacy Planner, a web-based decision-support framework that combines rule-based financial logic with machine learning–assisted analysis to evaluate property affordability and long-term financial feasibility. The proposed system enables users to construct a structured financial profile, analyze savings potential, estimate loan obligations, and assess affordability through transparent scoring mechanisms. Unlike black-box financial tools, the framework emphasizes explainability by enforcing interpretable financial constraints alongside data-driven predictions. The system is implemented using a modular web architecture and demonstrates how hybrid intelligence can improve clarity and reliability in personal financial decision-making. Experimental evaluation through simulated user scenarios indicates that the proposed approach reduces calculation errors, improves financial awareness, and supports realistic goal planning. The framework highlights the role of explainable AI in consumer-centric financial applications and provides a scalable foundation for intelligent financial planning systems.

Keywords: Financial Planning, Property Affordability, Hybrid Intelligence, Explainable AI, Decision Support Systems

Abstract: Accessing timely preliminary medical guidance remains a critical challenge for individuals worldwide, with many relying on traditional healthcare systems that lack accessibility, personalization, and immediate response mechanisms. Current symptom checker applications provide limited diagnostic accuracy, lack real-time emergency detection capabilities, and fail to deliver comprehensive medical knowledge integrated with evidence-based treatment recommendations. The absence of intelligent, multi-agent healthcare systems leaves individuals underprepared to assess their health conditions, identify emergencies, and make informed decisions about seeking professional medical care. To address these limitations, the MediAssist AI platform integrates Artificial Intelligence, Machine Learning, and Multi-Agent System architecture to deliver personalized, interactive preliminary medical guidance at scale. The system leverages advanced Natural Language Processing and machine learning classifiers to analyze symptom descriptions and predict likely medical conditions based on user input. Real-time emergency detection scans for critical keywords across eight life-threatening categories, ensuring immediate safety guidance delivery within 500 milliseconds. AI-powered knowledge agents retrieve comprehensive medical information from structured databases, while treatment recommendation agents provide evidence-based care guidance with appropriate medical disclaimers. Through a user-centric web platform built with Streamlit and Flask, users describe symptoms naturally, receive instant emergency alerts when critical conditions are detected, view top-three condition predictions with confidence scores, access detailed medical knowledge about identified conditions, and obtain treatment recommendations with home care instructions. Stakeholders including general users, healthcare administrators, and medical professionals benefit from structured preliminary assessment workflows and comprehensive health literacy resources. By combining rule-based emergency triage with machine learning-powered symptom analysis and knowledge-driven medical guidance, the proposed solution significantly improves healthcare accessibility, reduces response time for critical situations, and enhances informed decision-making while democratizing access to preliminary medical consultation.

Abstract: Time-series forecasting plays a vital role in domains such as energy management, finance, healthcare, and smart infrastructure. This paper presents a deep learning–based framework for short-term electricity consumption forecasting using historical power usage data. The proposed system evaluates and compares three advanced architectures: Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), and Transformer-based models. A multivariate forecasting approach is adopted by incorporating power-related and calendar-based features. A sliding window strategy is used to predict future values from recent observations. The system is deployed through a Flask-based web interface that enables interactive visualization and real-time prediction. Experimental results indicate that the CNN model achieves the best performance for short-term forecasting, outperforming LSTM and Transformer models in terms of RMSE, MAE, and R² score. The proposed framework demonstrates the practical applicability of deep learning techniques for intelligent energy demand prediction.

Keywords: Time-Series Forecasting, Deep Learning, LSTM, CNN, Transformer, Energy Consumption, Flask

Abstract: This study presents the development of an advanced deep learning–based system for the automated detection of synovial sarcoma from microscopic soft tissue images using Convolutional Neural Networks (CNNs). Synovial sarcoma is a rare and highly aggressive subtype of soft tissue sarcoma, where accurate and early diagnosis is essential for effective clinical management and improved patient prognosis. The proposed CNN framework is designed to automatically extract and learn discriminative features from histopathological images, enabling reliable identification of patterns characteristic of synovial sarcoma. By minimizing manual interpretation and inter-observer variability, the system serves as a supportive diagnostic tool for pathologists, enhancing diagnostic accuracy and efficiency. The results demonstrate the potential of deep learning techniques in improving histopathological analysis and contributing to timely and precise cancer diagnosis.

Abstract: The integrity of academic assessment relies on the secure distribution of examination question papers, yet traditional centralized systems remain vulnerable to unauthorized access and premature leaks. This paper proposes a decentralized framework that integrates the Ethereum blockchain implemented via Ganache and IPFS to create a tamper-proof, transparent environment for the examination lifecycle. By leveraging AES-128 encryption, question papers are secured locally before being hosted on a decentralized storage layer, eliminating central points of failure. A core contribution of this work is the implementation of a Smart Contract-driven Time-lock mechanism, which programmatically enforces access control by barring the decryption of materials until a specific, verifiable timestamp is reached. Developed using Python Flask and Web3.py, the results demonstrate a robust, auditable, and role-based protocol that effectively mitigates the risk of early disclosure and ensures the immutable governance of sensitive academic data within a private blockchain environment.

Keywords: Secure Examination Workflow, Blockchain, Ganache, IPFS, Smart Contracts, Time-Lock Mechanism, AES-128 Encryption, Decentralized Storage, Web3.py.

Abstract: Accurate prediction of agricultural commodity prices is a critical challenge due to the highly dynamic and nonlinear nature of agricultural markets. Crop prices are influenced by multiple interdependent factors such as climatic variations, seasonal demand, production levels, supply chain disruptions, government policies, and global trade dynamics. Traditional forecasting techniques, including linear regression and classical time-series models such as ARIMA, often fail to model these complex interactions and long-term temporal dependencies, leading to limited prediction accuracy and poor adaptability under volatile market conditions. This paper proposes a hybrid crop price prediction framework that integrates Machine Learning and Deep Learning techniques to achieve reliable and long-term agricultural price forecasting. The proposed system combines the strengths of Extreme Gradient Boosting (XGBoost) and Long Short-Term Memory (LSTM) networks. XGBoost is employed to effectively model structured features and nonlinear relationships among economic, seasonal, and meteorological variables, while LSTM networks are utilized to capture long-term sequential dependencies and temporal trends in historical crop price data. An ensemble strategy is applied to merge predictions from both models, enhancing robustness and reducing forecasting error. The system utilizes historical market prices along with weather-related parameters such as temperature, rainfall, and humidity to generate price forecasts for multiple crops and market locations up to twelve months in advance. The proposed framework is implemented using a scalable web-based architecture, featuring a React-based interactive dashboard for visualization and a FastAPI-powered backend for efficient data processing and real-time prediction. Experimental evaluation using standard performance metrics including RMSE, MAE, MAPE, and R² score demonstrates that the hybrid ensemble model consistently outperforms individual machine learning and deep learning models.

Abstract: Fruit maturity grading plays a crucial role in agricultural quality control, supply chain management, and food processing industries. Traditional manual grading methods are subjective, time-consuming, and prone to human error due to variations in lighting conditions, fatigue, and individual perception. This paper presents an automated fruit detection and three-stage maturity grading system using deep learning and image processing techniques. The proposed system classifies fruits into three maturity stages—unripe, ripe, and overripe -by analyzing visual features such as color, texture, and surface patterns. A Convolutional Neural Network (CNN) model is trained using the Fruits-360 dataset, enhanced with additional maturity-stage images. Image preprocessing techniques including resizing, normalization, background removal, and noise reduction are applied to improve classification accuracy. Experimental results demonstrate that the proposed system achieves high accuracy and consistency, significantly reducing dependence on manual inspection. The system provides a scalable and efficient solution for intelligent agricultural applications

Keywords: Fruit Detection, Maturity Grading, Convolutional Neural Networks, Image Processing, Deep Learning, Smart Agriculture.

Abstract: This paper presents a web-based Multi-Factor Authentication System designed to improve security and prevent unauthorized access to web applications. The system verifies user identity using multiple authentication layers such as password-based login, email-based One-Time Password (OTP) verification, and face recognition. By combining these security mechanisms, the system reduces the risks associated with traditional single-factor authentication methods. The proposed system integrates secure authentication logic with a modern web application to provide real-time verification at each stage of the login process.In addition to authentication, the platform includes features such as user registration, login activity tracking, session management, and administrative monitoring, making it suitable for real-world security applications. This approach demonstrates how multi-factor authentication techniques can deliver a reliable,scalable, and user-friendly solution for strengthening web application security.

Keywords: Multi-Factor Authentication, Web Security, One-Time Password, Face Recognition,Session Management, Secure Web Application.

Abstract: This paper presents a Real-Time Chat Application developed using the MERN stack and Socket.IO, designed to enable instant and reliable communication between users. The system facilitates real-time message exchange by establishing persistent bidirectional connections, allowing users to send and receive messages without page refresh or noticeable delay. The application supports essential chat features such as user authentication, one-to-one messaging, real-time message delivery, and online user status tracking. The backend of the system is implemented using Node.js and Express.js, while MongoDB is used for secure and efficient data storage. Socket.IO is integrated to handle real-time communication and event-based message broadcasting. The frontend is developed using React.js, providing a responsive and interactive user interface. Additional features such as message history storage, timestamp display, and user session management enhance the practicality of the application for real-world usage. The proposed solution demonstrates how modern web technologies and real-time communication frameworks can be combined to create a scalable, efficient, and user-friendly chat platform, suitable for instant messaging applications and collaborative environments.

Keywords: Real-Time Chat Application, MERN Stack, Socket.IO, WebSockets, Instant Messaging, Full Stack Development

Abstract: Traditional manual attendance systems in academic institutions suffer from operational deficiencies, temporal consumption, and vulnerability to "proxy" attendance. This paper proposes the Smart Facial Recognition Attendance System, an interaction-free biometric solution leveraging computer vision and machine learning. The system utilizes MediaPipe for high-precision extraction of 468 facial coordinates, ensuring robustness against cranial orientation fluctuations. A Random Forest Classifier is employed for identity categorization based on extracted facial embeddings, achieving a recognition precision surpassing 95% under regulated conditions. Integrated via a Flask web architecture and SQLite database, the framework provides real-time monitoring and automated report generation in CSV format. Experimental results indicate a significant reduction in administrative overhead and enhanced data integrity for institutional governance.

Keywords: Facial Recognition, MediaPipe, 468 Facial Landmarks, Random Forest Classifier, Automated Attendance System, Flask Web Framework, Computer Vision, Biometric Authentication, Real-time Monitoring, Machine Learning

Abstract: The proliferation of counterfeit merchandise in the global retail market represents a critical economic challenge, undermining brand integrity and exposing consumers to inferior, unregulated products. Traditional methods of authentication—often reliant on manual inspection by human experts—suffer from a "scalability-accuracy" bottleneck, rendering them inefficient for high-volume supply chains and e-commerce platforms. Furthermore, conventional image classification models frequently struggle with the nuanced, localized distortions typical of high-quality "super-fakes," such as incorrect font kerning or minor geometric deviations. This research introduces "Logo LIES" (Logo Identification & Estimation System), an integrated forensic framework that unifies real-time object detection with an automated verification pipeline. The system bypasses the limitations of standard two-stage detectors by adopting the Ultralytics YOLOv8 architecture, a single-shot regression model optimized for speed and precision. By utilizing a custom-trained dataset of authentic and counterfeit brand marks, the framework achieves high-fidelity localization of logo anomalies in sub-second inference times. To resolve the challenge of user accessibility, the system implements a "Liquid" User Interface coupled with an AI-driven forensic chatbot powered by the Google Gemini API. This architecture is specifically engineered to abstract complex neural network outputs into actionable, plain-language advice. A defining innovation of this project is its dual-modality inference engine, which facilitates both live webcam scanning and static high-resolution image analysis through a shared Flask middleware. Empirical testing confirms that the proposed system delivers a robust, low-latency solution capable of operating on standard consumer hardware without dedicated GPU acceleration. By democratizing access to advanced brand protection tools, this work contributes to the development of a secure digital retail ecosystem.

Abstract: The digital transformation of environmental monitoring represents a critical frontier in computational sustainability, with profound implications for public health, agricultural efficiency, and smart city infrastructure. Traditional methodologies for water quality assessment frequently encounter a "logistical-latency" bottleneck, where high-fidelity laboratory analysis requires significant time and specialized personnel, rendering real-time safety verification for rural and remote areas impractical. Furthermore, conventional manual testing is often compromised by human error, sample degradation during transport, and a lack of contextual data interpretation. This research introduces an Integrated Multi-Modal Prediction Framework that unifies Machine Learning classification, domain-specific rule engines, and Generative AI assistance into a singular, high-performance ecosystem. The system bypasses the limitations of standard "black-box" predictors by adopting a hybrid decision-making approach. By utilizing a Random Forest Classifier alongside a deterministic rule set, the framework achieves high-fidelity assessment of water potability based on 9 critical physicochemical parameters. This transformation of raw chemical data into actionable safety verdicts allows for instant execution without the necessity for expensive laboratory infrastructure. To resolve the challenge of interpreting complex chemical interactions, the system implements a Large Language Model (LLM) interface via the Google Gemini API. This architecture is specifically engineered to model the context of user queries, enabling the system to distinguish between "safe for agriculture" and "safe for drinking" scenarios. A defining innovation of this project is its decoupled modular architecture, which facilitates the independent execution of prediction, mapping, and advisory modules through a shared, optimized backend stream. The integration of geospatial visualization and automated history logging further ensures the utility of the platform for long-term monitoring. Empirical testing confirms that the proposed system delivers a robust, low-latency solution capable of operating on standard web servers. By democratizing access to advanced water safety analysis, this work contributes to the development of inclusive technology that bridges the gap between complex environmental data and public understanding.

Abstract: The increasing volume of written assessments in education has made manual essay evaluation time-consuming, subjective, and inconsistent. To address these challenges, this paper presents an Automated Essay Grading System that uses Natural Language Processing (NLP) and Machine Learning techniques to evaluate essays efficiently and fairly. The proposed system combines TF-IDF features and BERT-based embeddings to capture both statistical and contextual meaning of text, and employs a Ridge Regression model for accurate score prediction. In addition to grading, the system provides detailed feedback including grammar checking, readability analysis, sentiment evaluation, and bias detection, helping students improve their writing skills. A user-friendly web interface developed using Streamlit allows users to submit essays and view results instantly. The system also supports multilingual essay grading for regional languages such as English, Kannada, Telugu, and Tamil, making it more inclusive and accessible. Experimental results show that the system delivers consistent, scalable, and objective evaluation while significantly reducing grading effort. This work demonstrates the potential of AI-driven assessment tools in enhancing modern educational practices.

Keywords: Automated Essay Grading, Natural Language Processing, Machine Learning, BERT, TF-IDF, Ridge Regression, Multilingual Essay Evaluation, Educational Technology, AI-Based Assessment

Abstract: Early screening of cognitive and neurodevelopmental disorders remains a significant challenge due to delayed clinical access, limited awareness, and lack of scalable assessment tools. This work presents an AI-Driven Cognitive and Neurodevelopmental Risk Assessment Platform aimed at supporting preliminary screening of Autism Spectrum Disorder (ASD), Attention Deficit Hyperactivity Disorder (ADHD), and Dementia across different age groups. The system adopts a structured three-level assessment framework. Level-1 performs age-specific questionnaire-based screening using supervised machine learning models to estimate risk probability. Users exceeding a predefined threshold are directed to Level-2, which employs interactive and gamified cognitive tasks to capture behavioural and attention-related indicators. Level-3 provides alert-based guidance and professional consultation recommendations for high-risk cases. The platform integrates modern web technologies, secure backend services, and conversational AI using the Gemini API to ensure explainability and user engagement. Experimental evaluation demonstrates reliable age-adaptive interface behaviour, consistent risk prediction, and controlled assessment progression. The proposed system functions as an ethical, scalable early screening and decision-support tool that bridges the gap between self-assessment and clinical evaluation.

Abstract: Traditional manual study methods in academic environments suffer from operational deficiencies, temporal consumption, and lack of personalization. This paper proposes the AI-Based Study Assistant, an intelligent educational platform leveraging natural language processing and machine learning. The system utilizes NLTK and spaCy for automated text summarization and question generation from uploaded study materials. A Flask web architecture integrated with SQLite database provides real-time quiz generation and performance analytics. The framework automatically creates flashcards, summaries, and multiple-choice questions from PDF, DOCX, and TXT files, achieving contextual relevance exceeding 92%. An intelligent reminder system operates via background threading, sending email notifications for consistent study habits. Experimental results indicate significant reduction in study preparation overhead and enhanced learning retention through personalized assessment generation.

Keywords: Artificial Intelligence, Natural Language Processing, Intelligent Tutoring Systems, Automated Quiz Generation, Personalized Learning, Flask Framework, Educational Technology, Machine Learning, Study Management, Performance Analytics

Abstract: AirQ: Intelligent Air Quality Prediction and Alerting System is designed to monitor, analyze, and predict air pollution levels using environmental data. The system collects parameters such as PM2.5, PM10, temperature, humidity, and harmful gases to evaluate air quality conditions. Machine learning techniques are applied to forecast future pollution levels and identify potential health risks. Based on predicted air quality index (AQI) values, the system generates real-time alerts to inform users about unsafe conditions. This proactive approach helps individuals and authorities take preventive measures, supports environmental awareness, and promotes public health protection through data-driven decision-making.

Keywords: Air Quality Index, Pollution Prediction, Machine Learning Models, Environmental Monitoring, Health Alert System

Abstract: Human computer interaction is commonly performed using physical input devices, which may not be suitable in all environments or for all users. This paper presents Air Command, a vision-driven gesture and gaze control system that enables hands-free computer interaction using a standard webcam. The system combines gaze-based cursor movement with hand gesture-based command execution, allowing users to perform operations such as clicking, scrolling, zooming, and application control without physical contact. Eye blink detection is used for mouse click actions, while predefined hand gestures trigger system commands. The proposed approach operates in real time on consumer-grade hardware and incorporates stability mechanisms to reduce unintended actions. Experimental results demonstrate reliable performance under normal conditions, highlighting the feasibility of vision-based interaction as an effective alternative to traditional input devices.

Keywords: Human–Computer Interaction, Computer Vision, Gaze Tracking, Hand Gesture Recognition, Touch-Free Interaction, Assistive Computing.

Abstract: Lung cancer remains a leading cause of cancer-related mortality, where early diagnosis is critical for improving survival outcomes. Computed Tomography (CT) imaging is commonly used for lung cancer screening; however, manual interpretation of CT scans is time-consuming and susceptible to diagnostic variability. This paper presents LUNG VISION, an automated lung cancer detection and classification system based on machine learning and deep learning techniques. The proposed framework includes image preprocessing, lung region segmentation, feature extraction, and classification. Preprocessing techniques such as resizing, normalization, and noise reduction are applied to enhance CT image quality. Machine learning classifiers including Decision Tree, Random Forest, and Gaussian Naive Bayes are implemented using Histogram of Oriented Gradients features. In parallel, deep learning models such as Convolutional Neural Networks, DenseNet, and ResNet are employed through transfer learning to automatically learn discriminative features from CT images. The system classifies CT scans into normal, benign, and malignant categories and provides severity-related insights to support clinical decision-making. Experimental results indicate that deep learning models achieve superior diagnostic accuracy and robustness compared to traditional machine learning methods. The system is deployed via a web-based interface to assist radiologists in early and reliable lung cancer diagnosis.

Keywords: Lung Cancer Detection, Computed Tomography (CT), Deep Learning, Convolutional Neural Network (CNN), Machine Learning.

Abstract: Road accidents and vehicle breakdowns are increasing rapidly due to the continuous growth in vehicle usage and urban traffic density, resulting in frequent delays in emergency assistance and roadside services. In many critical situations, the absence of accurate location information, combined with manual communication methods, significantly delays rescue operations, which may lead to serious injuries, loss of life, and property damage. To overcome these challenges, this paper presents an Intelligent Road Rescue Application, a real-time, location-based emergency assistance system designed to provide fast, reliable, and transparent roadside rescue services. The proposed system connects users facing emergencies with nearby service providers such as mechanics, towing services, fuel suppliers, and ambulances through a centralized digital platform. By leveraging GPS technology, cloud computing, and real-time communication frameworks, the application enables automatic location detection, nearby vendor identification, continuous live tracking, secure digital payments, and structured feedback mechanisms. The system significantly improves response time, enhances transparency, and ensures efficient coordination between users and service providers. This paper presents a detailed discussion of the system architecture, methodology, implementation, simulation framework, and performance evaluation of the proposed solution.

Keywords: Roadside Assistance, GPS Tracking, Real-Time Systems, Emergency Services, Web Application

Abstract: The financial markets are characterized by high volatility, where retail traders often suffer significant capital losses due to emotional decision-making, cognitive bias, and inadequate risk management strategies. This paper presents TradeAi Pro, a comprehensive web-based algorithmic trading support system designed to democratize institutional-grade market analysis. By combining Computer Vision (OCR) for automated asset recognition with a Hybrid Machine Learning architecture (integrating XGBoost classifiers and Long Short-Term Memory neural networks), the system bridges the gap between raw market data and actionable trading insights. A distinguishing feature of the platform is its automated "Risk Logic Engine," which strictly enforces a 1:2 Risk-to-Reward ratio by dynamically calculating Stop Loss and Take Profit levels based on the asset's Average True Range (ATR). Furthermore, the application includes an interactive AI Trading Coach and an automated journaling module. Ultimately, this framework ensures that trading decisions are data-driven, mathematically sound, and minimized for psychological bias.

Keywords: Algorithmic Trading, Hybrid AI, XGBoost, LSTM, Computer Vision, Risk Management, FinTech.

Abstract: Modern educational institutions face increasing workloads due to repetitive academic tasks related to attendance management, assignment handling, grade processing, and communication among students, teachers, and parents. Traditional systems such as manual registers, spreadsheets, and fragmented digital tools are inefficient and fail to provide real-time updates, resulting in delays, errors, and reduced transparency. Although digital portals exist, many lack proper integration, real-time communication, and strong role-based access control required for large-scale deployment. This paper presents SmartEduConnect, a web-based academic management and communication platform that delivers centralized, role-based, and real-time academic services through a unified interface. The system integrates a modern full-stack architecture using React.js, Node.js, Express.js, and MongoDB, with Socket.io for real-time notifications and chat. SmartEduConnect supports administrators, teachers, students, and parents with personalized dashboards for managing classes, attendance, assignments, grades, and communication. Secure access is ensured through JWT-based authentication and role-based authorization. Experimental usage demonstrates reduced administrative workload, faster academic updates, improved data accuracy, and enhanced communication, validating the effectiveness of SmartEduConnect as a scalable and secure solution for academic management.

Keywords: Academic Management System, Educational Technology, Role-Based Access Control, Real-Time Communication, MERN Stack, Web Application, Student Information System

Abstract: In today’s highly competitive job market, organizations receive an overwhelming number of resumes for each job opening, making manual resume screening inefficient, time-consuming, and prone to human bias. Traditional recruitment methods rely heavily on keyword-based filtering and manual shortlisting, which often leads to inconsistent evaluations and missed opportunities to identify suitable candidates. To address these challenges, this paper presents an AI-Powered Resume Analyzer, an intelligent recruitment support system that automates resume parsing, skill extraction, candidate evaluation, and job–role matching using Artificial Intelligence (AI), Natural Language Processing (NLP), and Machine Learning (ML) techniques. The proposed system analyzes resumes uploaded in digital formats and extracts structured information such as skills, education, experience, certifications, and personal details. Machine learning models evaluate resumes against predefined criteria and job requirements to generate suitability scores, rank candidates, and provide actionable insights such as skill gaps and improvement recommendations. The system is developed using Python, NLP libraries, machine learning frameworks, and a web-based interface, ensuring scalability, accuracy, and usability. Experimental evaluation demonstrates that the system significantly reduces resume screening time while improving recruitment accuracy and fairness. The AI Resume Analyzer offers a practical, scalable, and efficient solution for modern recruitment automation.

Keywords: AI Resume Analyzer, Recruitment Automation, Natural Language Processing, Machine Learning, Resume Parsing, Intelligent Hiring Systems

Abstract: The secure exchange of textual information has become a critical requirement due to the rapid growth of online communication and data sharing. Text encryption plays a vital role in protecting sensitive information from unauthorized access, interception, and misuse. This project focuses on the design and implementation of a text encryption system that converts readable text into an unreadable format using cryptographic techniques, ensuring data confidentiality during storage and transmission. The proposed system applies an encryption algorithm to transform plain text into cipher text using a secret key, making the content accessible only to authorized users who possess the corresponding decryption key. The system also supports accurate decryption, restoring the original message without data loss. Emphasis is placed on simplicity, efficiency, and reliability so that the encryption process can be easily integrated into real-world applications. By implementing this text encryption mechanism, the project demonstrates how cryptographic methods can effectively safeguard personal, academic, and organizational data. The solution enhances security awareness and highlights the importance of encryption in preventing data breaches and maintaining privacy in modern communication systems.

Keywords: Text Encryption, Data Security, Cryptography, Plain Text, Cipher Text, Encryption Key, Decryption Process, Secure Communication, Information Privacy, Cyber Security

Abstract: In the digital era, users often provide consent on online platforms without fully understanding the implications of privacy policies and data-sharing terms. The Digital Consent Tracker aims to solve this issue using Artificial Intelligence (AI) and Natural Language Processing (NLP). The system extracts key points from lengthy privacy agreements, generates userfriendly summaries, and categorizes them into areas such as data collection, thirdparty sharing, data retention, and user rights. It offers a web dashboard and browser extension to help users instantly review simplified versions of consent documents. This approach promotes transparency, improves user awareness, and encourages ethical data-handling practices. Overall, this project demonstrates how AI-driven solutions can support informed decision-making, build user trust, and align with global privacy regulations such as GDPR and CCPA.

Keywords: AI, NLP, Privacy Policy, Digital Consent, Data Protection, GDPR, CCPA, Summarization, Browser Extension, User Awareness.

Abstract: The healthcare sector is currently undergoing a paradigm shift from reactive treatment to proactive, predictive care, driven by the explosion of "Big Data" from Electronic Health Records (EHRs) and wearable devices. However, the integration of this data into daily clinical practice remains fragmented, leading to reliance on manual diagnostics that are error-prone and time-consuming. This paper presents the Smart Healthcare Analysis System, a web-based predictive modeling framework designed to bridge the gap between raw medical data and actionable clinical insights. The system introduces a five-stage architectural pipeline combining advanced data preprocessing (including SMOTE for class imbalance), robust machine learning classification (utilizing Random Forest and XGBoost), and explainable AI techniques. Evaluated against standard healthcare datasets, the system achieves a predictive accuracy of over 90% in disease risk assessment while providing real-time decision support (<2 seconds latency). Unique to this framework is the integration of Feature Importance Analysis, which enhances clinical trust by transparently visualizing the physiological parameters driving each prediction. This work offers a scalable, economically feasible solution for modernizing healthcare delivery, particularly in resource-constrained environments.

Keywords: Healthcare Analytics, Machine Learning, Predictive Modeling, Decision Support Systems, Explainable AI, Disease Prediction.

Abstract: The growing demand for digital content creation and image editing has increased the need for intelligent systems that can automate these tasks efficiently. This project presents an AI-Powered Content and Image Enhancement Suite, a unified web-based platform that leverages Artificial Intelligence to generate, enhance, and refine textual and visual content. The system integrates Natural Language Processing (NLP) models for tasks such as article writing, blog title generation, and resume review, along with Computer Vision (CV) models for AI image generation, background removal, and object removal. The proposed solution enables users to perform multiple content enhancement operations through a single dashboard. By automating creative and editing processes, the system improves productivity, ensures consistency in output quality, and reduces manual effort. Experimental evaluation confirms the effectiveness of the platform in generating accurate, visually appealing, and contextually relevant results across all modules.

Keywords: Artificial Intelligence, Natural Language Processing, Computer Vision, Content Generation, Image Enhancement, Web Application.

Abstract: Cardiovascular diseases are among the leading causes of global mortality, making early prediction and preventive diagnosis a critical requirement in modern healthcare systems. This paper presents an artificial intelligence and machine learning–based approach for predicting cardiovascular disease using patient clinical and lifestyle data. The proposed system applies supervised machine learning algorithms to analyze key medical attributes such as age, blood pressure, cholesterol levels, blood glucose, heart rate, and behavioral factors. Effective data preprocessing techniques, including data cleaning, normalization, and feature selection, are employed to enhance model accuracy and reliability. The trained model identifies complex patterns and relationships within the dataset to classify individuals based on cardiovascular risk levels. Experimental results demonstrate that the proposed approach achieves improved prediction performance compared to traditional diagnostic methods. The system provides a scalable, cost-effective, and decision-supportive solution that can assist healthcare professionals in early detection and risk assessment of cardiovascular diseases.

Keywords: Cardiovascular Disease Prediction, Artificial Intelligence, Machine Learning, Supervised Learning, Medical Data Analysis, Risk Assessment, Healthcare Analytics.

Abstract: The rapid growth of online shopping has increased the demand for scalable, secure, and user-friendly e-commerce platforms. This paper presents ShopEase, a full-stack e-commerce web application developed using the MERN stack, which includes MongoDB, Express.js, React.js, and Node.js. The system enables users to browse products, manage shopping carts, and complete secure transactions through an integrated backend architecture. A key innovation of this project is the integration of a rule-based live chatbot, implemented using pure JavaScript, which provides real-time assistance to users by answering frequently asked questions, guiding navigation, and improving user engagement without relying on external AI services. Security is ensured using JSON Web Token (JWT) based authentication for safe user sessions and role-based access control. The results demonstrate that ShopEase offers an efficient, modular, and scalable solution suitable for modern e-commerce applications.

Keywords: E-Commerce, MERN Stack, React.js, Node.js, MongoDB, JWT Authentication, Rule-Based Chatbot

Abstract: With the rapid growth of digital communication, professionals receive a large volume of emails daily, making manual email drafting time-consuming and repetitive. To address this challenge, this paper presents a Smart Email Assistant that automatically generates context-aware, professional email replies using Natural Language Processing (NLP) and Artificial Intelligence (AI). The proposed system analyzes the original email content, identifies intent, tone, and key entities, and generates a suitable reply based on user-selected parameters such as tone and purpose. The application provides features including reply generation, history tracking, dashboard analytics, and one-click copy or download functionality. The system is implemented using React with Vite for the frontend and an AI-powered text generation engine for response creation. Experimental usage shows that the Smart Email Assistant significantly reduces response time, improves consistency, and enhances productivity for students, job seekers, and professionals. This work demonstrates how AI-driven email automation can streamline digital communication workflows.

Keywords: Smart Email Assistant, Automated Email Reply Generation, Natural Language Processing (NLP), Artificial Intelligence (AI), React and Productivity tools.

Abstract: Pancreatic cancer is one of the most aggressive and life-threatening malignancies due to its late diagnosis, complex progression patterns, and limited treatment options. Traditional diagnostic approaches primarily rely on radiological interpretation and clinical biomarkers, which are often subjective and insufficient for early detection. To address these challenges, this paper proposes an Explainable Multimodal Artificial Intelligence (AI) system for the prediction and detection of pancreatic cancer. The proposed system integrates CT/MRI medical imaging data with clinical and laboratory parameters to perform comprehensive cancer analysis. Advanced machine learning and deep learning techniques are employed to extract meaningful features from multimodal inputs, enabling accurate cancer stage prediction and survival estimation. Explainable AI (XAI) methods such as heatmaps and feature importance analysis are incorporated to enhance model transparency and clinical trust. The system is implemented using Python, Flask, and modern AI frameworks, providing a scalable web-based diagnostic platform. Experimental results demonstrate improved prediction accuracy, reduced diagnostic uncertainty, and enhanced interpretability, making the proposed system a reliable clinical decision-support tool.

Keywords: Pancreatic Cancer Prediction, Multimodal AI, Explainable AI, Medical Imaging, Clinical Data Analysis, Deep Learning

Abstract: In the contemporary software engineering landscape, developers and students frequently encounter the challenge of onboarding to large, unfamiliar codebases. Platforms like GitHub host millions of repositories, yet understanding the underlying logic, architecture, and dependency flow of these projects remains a labor-intensive process dependent on manual traversal and often outdated documentation. To mitigate this inefficiency, this paper presents the Agentic AutoCode Analyzer, a web-based intelligent system designed to automate the comprehension of software repositories. The proposed system accepts a GitHub repository URL, autonomously performs a shallow clone operation to minimize bandwidth usage, and recursively maps the directory structure to build a comprehensive context object. By integrating a Large Language Model (LLM) reasoning engine via a local or API-based inference layer, the system functions as an interactive "agent." This agent assists users by answering architectural queries, explaining specific code syntax, and summarizing project objectives. Experimental results indicate that the system significantly reduces the cognitive load required for code comprehension and offers a viable tool for both educational and professional software development environments.

Keywords: Artificial Intelligence, Static Code Analysis, Large Language Models (LLM), Software Engineering Education, Automated Documentation, React.js, Node.js.

Abstract: This project, titled CertChain: Blockchain-based Certificate Management System, addresses the challenges of digital certificate issuance, verification, and security. Traditional paper-based or centralized digital systems are prone to forgery, tampering, and mismanagement. CertChain leverages blockchain technology to ensure tamper-proof, verifiable, and transparent certificate records. Users can securely register, request, and download certificates, while administrators manage issuance and revocation.

Abstract: Mental health challenges have become increasingly common due to academic pressure, work stress, social isolation, and lifestyle changes. However, access to timely mental health support remains limited for many individuals because of cost, stigma, and lack of availability of professionals. This project, titled AI Powered Therapy – SafeSpace for Mental Health Support, aims to provide an accessible and supportive digital platform that assists users in managing their emotional well-being. Safe Space is an AI-based conversational system designed to offer empathetic responses, emotional guidance, and practical coping strategies through text-based interaction. The system analyzes user input to understand emotional states such as stress, anxiety, and loneliness, and responds in a supportive and non-judgmental manner inspired by basic cognitive behavioral therapy principles. The primary goal of the system is to provide immediate emotional support rather than replacing professional therapists. To enhance user safety, the project includes an emergency call module that is activated when severe emotional distress is detected. This feature uses the Twilio API to initiate calls to emergency contacts or helpline numbers, ensuring timely human intervention during critical situations. Additionally, the system provides nearby therapist contact details to encourage users to seek professional help when required. The Safe Space system follows a modular and user-friendly design, integrating chatbot interaction, emergency support, and therapist recommendations into a single platform. The project demonstrates how artificial intelligence can be responsibly used to support mental health awareness and early intervention, making emotional support more accessible while promoting professional care when needed. With the advancement of technology, especially Artificial Intelligence (AI), new opportunities have emerged to support mental health in a more accessible and user-friendly manner. AI-based systems can provide immediate responses, maintain privacy, and offer a safe environment where users feel comfortable expressing their emotions. These systems do not aim to replace human therapists but can act as a supportive companion, especially during moments when professional help is unavailable.

Abstract: Distributed Denial of Service (DDoS) attacks are one of the most common threats to modern web applications, causing service unavailability and performance degradation by overwhelming servers with excessive traffic. Early identification of such attack conditions is essential to reduce downtime and maintain service reliability. This paper presents DDoS Defender, a lightweight and ethical web-based system designed to detect potential DDoS attack scenarios using safe and non-intrusive website performance analysis.

Abstract: Visual impairment significantly restricts independent access to printed and digital textual information, affecting millions worldwide. Traditional assistive solutions suffer from high costs, limited language support, and dependence on specialized hardware. This paper presents a comprehensive multilingual OCR-based assistive system designed to enable visually impaired individuals to access textual content across linguistic boundaries. The system integrates optical character recognition, automated translation, and text-to-speech synthesis into a unified web-based platform supporting six languages including English and major Indian regional languages (Hindi, Tamil, Telugu, Malayalam, Kannada). Implementation utilizes FastAPI framework for asynchronous processing, Tesseract OCR engine configured with language-specific models, dual-fallback translation architecture ensuring 99.2% service availability, and Google Text-to-Speech for natural audio generation. Image preprocessing techniques including grayscale conversion and adaptive thresholding enhance recognition accuracy by 20-25%, particularly effective for complex Indic scripts. The system accepts images through web-based interface, processes them through sequential pipeline of enhancement, text extraction, translation, and speech synthesis, delivering comprehensive output within 4.6 seconds average processing time. Experimental results demonstrate 95% OCR accuracy for printed text across supported languages, successful translation with dual-fallback reliability, and high user satisfaction ratings (4.8/5 for ease of use, 4.7/5 for audio clarity, 4.6/5 overall). The framework addresses critical accessibility gaps in multilingual environments while maintaining cost-effectiveness through open-source libraries and web-based deployment accessible from standard computing devices.

Keywords: Optical Character Recognition, Assistive Technology, Multilingual Processing, Text-to-Speech Synthesis, Accessibility, Computer Vision, Visual Impairment, Natural Language Processing, Image Processing, Web-based Application

Abstract: With the rapid growth of online news platforms, users often face information overload due to the availability of large volumes of news content. Most existing news systems either provide generic news feeds or rely completely on automated recommendation techniques, which may not always reflect user intent. This paper presents News-Mania, a news recommendation system that combines manual personalization and AI-based personalization to deliver relevant news. In the proposed system, users manually select their preferred news categories, which ensures direct control over content selection. In addition, Artificial Intelligence analyzes user interaction data such as reading behavior and previously viewed articles to further refine recommendations. News articles are first fetched from online sources and stored in a database before being processed for recommendation. This hybrid approach improves recommendation accuracy, enhances user satisfaction, and reduces irrelevant content exposure.

Keywords: Artificial Intelligence, News Recommendation System, Manual Personalization, AI Personalization, User Preferences.

Abstract: Potato is one of the most important food crops worldwide, and its productivity is greatly affected by various plant diseases. Early and accurate detection of potato plant diseases is essential to reduce crop losses and improve agricultural yield. Traditional disease identification methods rely on manual observation by experts, which is time-consuming, costly, and prone to human error. This project presents a Potato Plant Disease Classification system using a Convolutional Neural Network (CNN) to automatically identify diseases from leaf images. The proposed model is trained on a dataset containing images of healthy potato leaves and diseased leaves such as Early Blight and Late Blight. Image preprocessing techniques like resizing and normalization are applied to enhance model performance. The CNN model extracts important features from leaf images and classifies them accurately into respective disease categories. Experimental results show that the CNN-based approach achieves high accuracy and efficiency in disease classification. This system can help farmers and agricultural professionals in early disease detection, enabling timely treatment and reducing crop damage. The proposed solution demonstrates the potential of deep learning techniques in smart agriculture and precision farming.

Abstract: The increasing number and scale of government welfare schemes have created a strong need for effective systems that can analyze scheme performance and improve transparency. One of the major challenges faced by administrators and policymakers is the lack of a centralized mechanism to monitor beneficiary coverage, financial utilization, and overall scheme impact. Traditional methods rely on manual reports and fragmented data sources, which often result in delayed insights and inefficient decision-making. The Government Scheme Analysis and Tracker is developed to address these challenges by providing an intelligent and automated platform for analyzing welfare schemes. It is a web-based application that processes scheme-related data to evaluate beneficiary trends, expenditure patterns, and district-wise and state-wise performance. By using analytical and forecasting techniques, the system enables users to understand how schemes are performing and identify areas that require improvement without relying on manual analysis. The system is implemented using React.js for the frontend interface, FastAPI for backend processing, MongoDB for data storage, and Python-based analytics and AI models for data analysis and forecasting. The project demonstrates effective integration of modern web technologies and artificial intelligence to support data-driven governance. The Government Scheme Analysis and Tracker aims to enhance transparency, support informed policy decisions, and improve the overall effectiveness of government welfare programs

Abstract: The increasing digitization of power grid infrastructure has introduced cybersecurity vulnerabilities requiring robust anomaly detection mechanisms. This study presents a large-scale empirical evaluation framework for characterizing the behavior of six unsupervised anomaly detection paradigms under diverse smart grid attack scenarios. The evaluation encompasses self-supervised contrastive learning with temporal convolutional encoders alongside established methods including Isolation Forest, One-Class Support Vector Machine, Autoencoder, Deep Support Vector Data Description, and Local Outlier Factor. Experiments were conducted across two complementary datasets comprising over 2.1 million power consumption records representing both synthetic perturbations and realistic attack scenarios with seven distinct threat types. Rather than identifying a universally optimal method, this study characterizes scenario-dependent performance patterns and operational trade-offs. Results demonstrate that all evaluated paradigms achieve Area Under the Receiver Operating Characteristic Curve values exceeding 0.90 on realistic attack scenarios, with F1 scores ranging from 0.637 to 0.806 depending on method and attack characteristics. The contrastive learning paradigm achieved F1 scores of 0.449 and 0.786 on synthetic and realistic scenarios respectively. An ablation study examining temporal augmentation strategies revealed marginal performance variations, suggesting that the learning objective rather than augmentation design drives representation quality. These findings establish reproducible benchmarks, characterize the strengths and limitations of each paradigm under different deployment conditions, and provide practical guidance for selecting anomaly detection approaches based on specific operational requirements rather than aggregate performance metrics. Index Terms: Smart grid security, Anomaly detection, Unsupervised learning, Empirical evaluation, Cybersecurity benchmarking, Critical infrastructure.

Abstract: Diabetes is a chronic and rapidly increasing health condition that requires early detection and continuous monitoring to prevent severe complications. Traditional diabetes screening methods often rely on manual evaluation or simple calculators that consider limited parameters, resulting in inaccurate or generalized risk assessment. To address these limitations, this project presents an AI-Powered Diabetes Prediction System that uses machine learning techniques to provide accurate, real-time, and personalized diabetes risk prediction.

Abstract: The rapid increase in lifestyle-related chronic diseases such as Type 2 diabetes, obesity, and hypercholesterolemia has created a need for intelligent, integrated, and personalized healthcare solutions. This paper presents the design and development of an AI Healthcare and Nutrition Assistance App, a comprehensive web-based platform that leverages artificial intelligence to provide data-driven nutrition guidance and continuous health monitoring. The system integrates AI-powered meal analysis using computer vision, personalized diet planning, and real-time health tracking to support preventive healthcare and disease reversal. The application follows a multi-role architecture comprising patients, doctors, dieticians, and administrators, each with role-based access and specialized dashboards. Core functionalities include an AI Meal Analyzer for calorie and nutrient estimation from food images, recipe recommendation systems, vitals tracking, clinical report uploads, and interactive data visualization of health metrics such as glucose levels and weight trends. Built using modern web technologies including Next.js, React, Firebase, and Google Generative AI, the platform ensures scalability, security, and usability. By bridging the gap between daily nutrition habits and clinical supervision, the proposed system enhances decision-making, promotes sustainable lifestyle changes, and demonstrates the potential of AI-driven digital health platforms in modern healthcare ecosystems.

Abstract: The increasing demand for short-term accommodation such as paying guest rooms and guest houses has highlighted the need for efficient and reliable digital booking platforms. Traditional guest room booking methods rely heavily on manual processes, phone calls, or informal communication, which often result in booking conflicts, data inconsistency, and poor customer experience. This paper presents the Guest Room Booking Application, a web-based system designed to automate and streamline the process of room listing, searching, booking, and management. The application is developed using modern web technologies following the MERN stack architecture, where React is used for the frontend, Node.js and Express.js for the backend, and MongoDB for database management. The system supports role-based access control, allowing room owners to manage rooms and bookings while enabling customers to search, view, and reserve rooms online. Secure authentication using JSON Web Tokens (JWT), real-time booking status updates, and integrated dashboards enhance system usability and security. Experimental results indicate that the proposed system significantly reduces booking errors, improves operational efficiency for room owners, and provides a smooth, user-friendly experience for customers. The Guest Room Booking Application offers a scalable and practical solution suitable for real-world deployment in small and medium-scale accommodation services.

Keywords: Guest Room Booking Application, Online Reservation System, MERN Stack, Web Application, Role- Based Access Control, MongoDB, React, Node.js

Abstract: The rapid growth of digital communication has increased the need for secure and reliable real-time video and chat applications with effective recording capabilities. Many existing platforms support live communication but lack integrated recording, organized storage, and easy retrieval of recorded sessions. This paper presents LiveNest, a web-based real-time video calls and chat application with built-in video call recording as a core feature. LiveNest enables users to securely authenticate, perform real-time messaging, conduct one-to-one and group video calls, share screens, and automatically record video sessions for future reference. The system is developed using React and Tailwind CSS for the frontend, Node.js and Express for backend services, and MongoDB for data storage. Real-time communication and recording are powered using Stream services to ensure low latency and scalability. Secure authentication is implemented using JSON Web Tokens (JWT). The modular architecture supports performance, security, and future enhancements, making LiveNest suitable for online collaboration, learning, and professional communication.

Abstract: With the increasing demand for intelligent and interactive software systems, virtual assistants have become an essential part of modern human–computer interaction. This paper presents SWARA – The Virtual Assistant, an AI-based desktop assistant designed to perform automated system tasks while also supporting structured interview practice. The proposed system operates in two distinct modes: General Assistant Mode, which executes voice- or text-based commands such as application control, web navigation, and system utilities; and Interview Mode, which simulates a mock interview environment by presenting questions, capturing user responses, and generating AI-driven feedback. The system is developed using Python and integrates speech recognition, natural language understanding, and AI response generation to enable smooth and natural interaction. A modular architecture is adopted to separately manage input handling, command interpretation, task execution, interview management, and response delivery, ensuring scalability and ease of maintenance. By combining productivity automation with interview-oriented learning support, SWARA provides a practical and interactive platform for improving communication skills and interview readiness. Experimental usage demonstrates that the assistant delivers responsive interaction, reliable task execution, and structured guidance, making it a useful intelligent tool for students and individual learners.

Keywords: Virtual Assistant, Artificial Intelligence, Speech Recognition, Desktop Automation, Interview Mode, Human–Computer Interaction.

Abstract: This paper presents a novel Adaptive E-Learning framework designed to enhance student engagement and content retention through hyper-personalization. By combining Large Language Models (LLMs) via the Groq API for dynamic study planning and the Tavily Search API for real-time resource curation, the system addresses the critical challenges of information overload and static curriculum delivery. This multi-modal approach integrates a Retrieval-Augmented Generation (RAG) pipeline, allowing students to interact with PDF textbooks contextually ("Chat with PDF") while ensuring high factual accuracy. The system features a "white-box" approach to content delivery, where every AI-generated answer is cited from the user's uploaded material. Additionally, the platform includes a dynamic resource curator that filters web content to reduce cognitive load, bridging the gap between open-ended internet search and structured academic learning.

Keywords: Adaptive Learning, Retrieval-Augmented Generation (RAG), Large Language Models (LLM), Personalized Education, Cognitive Load Management, MERN Stack.

Abstract: Emotional health has become an important concern in today’s fast-paced lifestyle, where individuals face continuous academic, professional, and social pressures. Many people hesitate to share their emotional struggles due to fear of being judged or due to limited access to immediate support. This paper presents WEHEAL, an AI-powered emotion-aware chatbot developed to provide a safe and private environment for emotional expression and self-reflection. The WEHEAL system applies Natural Language Processing techniques to analyze user text and identify emotional states such as stress, sadness, anxiety, anger, and happiness. Based on the identified emotion, the chatbot generates empathetic and context-sensitive responses. In addition to conversational support, the system includes mood tracking, journaling, daily reflection prompts, wellness exercises, and emotional analytics. The application is implemented using Python and Streamlit, with all user data stored locally in JSON format to ensure privacy and offline usability. The results indicate that WEHEAL effectively promotes emotional awareness and provides meaningful support, highlighting the potential of AI-based wellness applications.

Abstract: The increasing use of digital payments and group-based activities such as travel, shared accommodation, and events has created a strong need for efficient expense management systems. Many existing solutions provide basic expense tracking but lack intelligent automation, transparency, and user-friendly settlement mechanisms. This paper presents the AI Powered Expense Splitter, a web-based application designed to simplify shared expense tracking and settlement using automation and AI-assisted features. The system allows users to securely authenticate, create expense groups, add members, record expenses, and automatically split costs among participants. AI-based receipt processing reduces manual data entry by extracting expense details directly from uploaded receipts. The application provides clear balance summaries, showing how much each user owes or is owed. The system is developed using modern web technologies including Next.js, React, Tailwind CSS, Convex for backend and database management, and Clerk for secure authentication. The modular architecture ensures scalability, security, and ease of maintenance, making the system suitable for real-world personal and group expense management.

Abstract: Early detection of wheat diseases plays a vital role in minimizing crop losses and improving agricultural productivity. Wheat rust and leaf spot diseases are among the most common threats affecting wheat crops, and manual inspection methods are often time-consuming, subjective, and prone to delays. This project presents an automated image-based system for the detection and classification of wheat rust and leaf spot diseases using computational techniques. The system utilizes publicly available wheat leaf image datasets collected from online repositories and applies a structured workflow that includes image acquisition, preprocessing, feature extraction, classification, and prediction. Image preprocessing techniques such as resizing, noise removal, normalization, and enhancement are employed to improve image quality and highlight disease characteristics. Color and texture-based features are extracted from the preprocessed images to effectively distinguish between healthy and infected leaf samples. A machine learning classifier is trained using the extracted features to classify images into multiple disease categories, including healthy leaves. The trained model is evaluated using standard performance metrics, and experimental results demonstrate reliable classification accuracy, validating the effectiveness of the proposed approach. This system provides a dataset-driven solution that can assist in early disease identification and support timely decision-making for disease management. The proposed framework is scalable and can be extended to detect additional crop diseases in the future, contributing to the development of intelligent agricultural support systems.

Keywords: Wheat leaf analysis, Plant disease detection, Image enhancement, Feature-based classification, Crop health monitoring, image processing, Early disease identification

Abstract: Online communities often struggle with efficient networking due to incomplete information about members’ capabilities, interests, and intentions, making manual networking time-consuming and prone to mismatched partnerships. To overcome this challenge, the project implements a production-ready, AI-driven community networking platform that uses LLM-powered natural language search to intelligently match users based on comprehensive profile analysis, including Ikigai self-discovery responses, professional backgrounds, portfolios, social profiles, skills, and intent. The system features a structured multi-step onboarding process to capture purpose-driven and professional data, AI-assisted matchmaking powered by Google Gemini 2.5 Flash through the Lovable AI Gateway, and a natural language search interface that returns ranked and filterable match lists with percentage-based compatibility scoring, AI- generated match explanations, and highlighted attributes. In addition, the platform includes real-time built-in messaging with conversation threading, customizable privacy controls, and a modern gradient-themed user interface developed using Tailwind CSS and shadcn/ui components. The application is built on a modern, scalable technology stack comprising React 18.3, TypeScript, Vite for build tooling, Supabase for authentication and PostgreSQL database management, real-time subscriptions for messaging, Row Level Security (RLS) policies for secure data access, and a serverless edge-function architecture to efficiently support AI processing.

Abstract: Traffic rule enforcement in urban environments remains largely dependent on manual monitoring practices, leading to limited scalability, inconsistent evaluation, and delayed action against safety violations. One major challenge faced by transportation authorities is identifying two-wheeler riders who fail to wear helmets and subsequently generating actionable reports for follow-up. Existing surveillance systems provide limited automation and require human intervention to review video feeds, detect violations, and record offender details. This paper presents an intelligent Helmet Detection and Reporting System designed to automate the identification of helmet misuse and streamline violation tracking through computer vision techniques. The proposed framework utilizes deep learning–based object detection to locate motorcycles and analyze rider head regions from video footage, thereby determining helmet compliance in real time. In cases of detected violations, the system further extracts vehicle number plates, applies optical character recognition to identify registration numbers, and compiles structured violation evidence suitable for reporting. Unlike conventional manual workflows, the system provides continuous, data-driven, and unbiased assessment of helmet usage under varying traffic densities and lighting conditions. Experimental evaluation conducted on realistic traffic datasets demonstrates high detection accuracy, reduced dependency on manual supervision, and significant improvement in reporting efficiency. The proposed system highlights the potential of automated computer vision pipelines to enhance safety, support enforcement agencies, and promote rule compliance in intelligent transportation environments.

Keywords: Helmet Detection, Traffic Surveillance, Number Plate Recognition, Computer Vision, Deep Learning, OCR, Intelligent Transportation Systems

Abstract: The rapid digitization of photography has led to a fragmentation of family memories across disparate platforms, ranging from instant messaging apps that degrade quality to public social media networks that compromise privacy. Traditional cloud storage solutions, while secure, often present significant technical barriers for non-digital natives within a family unit. This paper presents the Family Gallery Management System, a secure, full-stack web application designed to centralize family media archives. The proposed system utilizes the MERN stack (MongoDB, Express.js, React.js, Node.js) to create a private social network governed by strict Role-Based Access Control (RBAC). A unique algorithmic approach to storage management enforces a strict 3GB resource limit per gallery, simulating enterprise resource constraints. Furthermore, the system introduces a password-less authentication mechanism based on phone numbers to ensure cross-generational accessibility. Experimental validation demonstrates that this architecture successfully balances the security of private cloud storage with the ease of use required for family social interaction, providing a robust solution for digital heritage preservation

Keywords: Web Development, MERN Stack, Role-Based Access Control, Cloud Storage, Family Social Network, Digital Archiving.

Abstract: In recent years, the rapid growth of interconnected digital systems has significantly increased the number and complexity of cyber attacks. Security Information and Event Management (SIEM) systems play an important role in monitoring and analyzing security events in modern organizations. However, traditional SIEM platforms rely mainly on static rule-based detection and manual analysis, which limits their ability to detect unknown threats and respond efficiently in real time. This paper proposes an AI-driven SIEM system for real-time cyber threat detection and security monitoring. The proposed system integrates network traffic analysis, system performance monitoring, and deep learning–based intrusion detection to provide intelligent and automated security analysis. A Convolutional Neural Network is used to classify network behavior as normal or suspicious, while an AI-based alert interpretation module generates concise, human-readable security summaries. The system also monitors CPU usage, memory consumption, and disk activity to provide holistic situational awareness. Experimental results show that the proposed system improves detection accuracy and reduces false alerts compared to conventional SIEM approaches. The developed framework offers an effective and scalable solution for modern cybersecurity environments.

Keywords: Security Information and Event Management, Intrusion Detection, Artificial Intelligence, Deep Learning, Cybersecurity, Real-Time Monitoring.

Abstract: Alzheimer’s disease is a progressive neurodegenerative disorder that leads to irreversible cognitive decline and memory impairment. Early detection of the disease is essential for effective treatment planning and improved patient care. Magnetic Resonance Imaging (MRI) plays a vital role in identifying structural changes in the brain associated with Alzheimer’s disease. Recent advancements in deep learning have enabled automated and accurate analysis of medical images, reducing dependency on manual interpretation. This paper presents a Deep Learning Framework for Alzheimer’s Disease detection using Brain MRI images. A Convolutional Neural Network (CNN) is developed to classify MRI images into different stages of Alzheimer’s disease after preprocessing techniques such as resizing and normalization. In addition to image-based classification, a supporting clinical data-based prediction module using a Random Forest classifier is integrated to enhance prediction reliability. The system is implemented as a web-based application that allows users to upload MRI images and clinical data for real-time prediction. Experimental evaluation demonstrates high classification accuracy and reliable performance, highlighting the effectiveness of combining deep learning and machine learning techniques for early Alzheimer’s disease detection. The proposed framework provides a scalable and practical solution suitable for academic research and future healthcare applications.

Keywords: Alzheimer’s Disease, Brain MRI, Deep Learning, Convolutional Neural Network, Random Forest, Medical Image Analysis

Abstract: Agricultural productivity is increasingly affected by unpredictable weather conditions, plant diseases, inefficient irrigation practices, and limited access to timely market and advisory information. These challenges are more prominent among small and medium-scale farmers who rely on traditional decision-making methods. To address these issues, this research presents an AI-Based Precision Agriculture Support System that provides comprehensive and intelligent assistance for modern farming. The proposed system integrates multiple analytical modules, including crop recommendation, crop yield prediction, plant disease detection, irrigation planning, market price analysis, and weather forecasting. Machine learning algorithms are employed to analyze soil characteristics, climatic parameters, historical crop data, and market trends, while deep learning techniques are used to identify plant diseases from leaf images at an early stage. The irrigation planning module utilizes predictive insights combined with weather forecasts to optimize water usage and reduce resource wastage. All modules are unified through a web-based platform that delivers real-time, user-friendly recommendations without dependence on complex sensing infrastructure. Experimental results demonstrate that the system produces reliable predictions and actionable insights, contributing to improved crop management, efficient resource utilization, and enhanced decision-making. The proposed solution offers a scalable, cost-effective, and sustainable approach toward intelligent agriculture and supports the adoption of data-driven farming practices.

Keywords: Precision Agriculture, Artificial Intelligence, Crop Recommendation, Yield Prediction, Plant Disease Detection, Smart Irrigation, Market Price Analysis, Weather Forecasting.

Abstract: Traffic congestion in urban areas poses significant challenges to emergency medical services, where delayed ambulance response times can result in preventable fatalities. Traditional traffic management systems operate on fixed timer-based schedules that fail to adapt dynamically to emergency situations, causing critical delays for ambulances navigating through congested roads. Manual intervention by traffic police is often inefficient and cannot scale across multiple intersections simultaneously. This project presents an innovative artificial intelligence-based solution for automated ambulance detection and emergency traffic clearance using YOLOv5 deep learning architecture. The proposed system leverages state-of-the-art computer vision techniques to detect ambulances in real-time from video feeds captured by traffic cameras or uploaded video files. The system processes visual data through advanced image preprocessing techniques and employs the YOLOv5 object detection algorithm to identify ambulances with confidence scores exceeding 0.5 threshold. The architecture comprises multiple integrated components including camera-based video capture, image preprocessing modules, the YOLOv5 detection engine, traffic signal control interfaces using NTCIP protocol, and comprehensive logging systems. The system supports both real-time video stream processing at 30 frames per second and batch processing of pre-recorded video files. Extensive testing demonstrates detection accuracy exceeding 97% under diverse lighting conditions and traffic scenarios. This cost-effective, scalable solution addresses critical limitations of existing hardware-based traffic management systems by providing a software-centric approach that can be deployed across urban infrastructure with minimal modifications to existing camera networks.

Keywords: Ambulance Detection, YOLOv5, Computer Vision, Traffic Signal Control, Deep Learning, Object Detection, Real-time Processing, Smart City, Emergency Response, Intelligent Transportation Systems

Abstract: Modern web applications demand rapid releases, consistent deployments, and strong security practices. Manual build and deployment approaches often lead to configuration drift, delayed delivery, and increased chances of human error. This paper presents the design and implementation of a CI/CD pipeline for a MERN-based e-commerce application using Jenkins and AWS, enhanced with DevSecOps practices. The proposed pipeline automates the complete workflow from source code integration to containerized deployment. It includes static code quality validation using SonarQube, vulnerability assessment using Trivy, container image packaging using Docker, private image management through AWS Elastic Container Registry (ECR), and deployment to AWS EC2. Additionally, monitoring and observability are improved using AWS CloudWatch and alerting mechanisms through Slack notifications. The implementation ensures faster and repeatable deployments, improved code reliability, and early identification of security issues, making the system suitable for real-world production workflows..

Keywords: CI/CD, DevOps, DevSecOps, Jenkins, Docker, SonarQube, Trivy, AWS EC2, AWS ECR, CloudWatch, MERN Stack, E-Commerce

Abstract: Conventional programming practice methods in educational environments face challenges such as delayed feedback, limited learner engagement, and restricted access to execution environments. This paper presents CodePlay, an interactive coding learning platform that integrates real-time code execution, animated algorithm visualization, and gamified problem-solving. The system provides an in-browser code editor connected to a backend compiler service, enabling instant compilation, execution, and error feedback. Step-by-step visual animations are incorporated to illustrate algorithm behavior, improving conceptual understanding. Gamified puzzle modules such as sorting challenges, graph coloring, and logic-based games enhance learner motivation and skill retention. Developed using a web-based architecture with JavaScript, Node.js, and lightweight backend services, the platform supports progress tracking and score-based evaluation. Experimental observations demonstrate improved learner engagement, faster error resolution, and enhanced programming proficiency through interactive and visual learning approaches.

Keywords: Interactive Coding Platform, Online Compiler, Algorithm Visualization, Gamification, Programming Education, Real-time Code Execution, Learning Analytics, Web-based Learning System, Skill Tracking, Computer Science Education

Abstract: This project presents a web-based AI-powered mental health analysis system designed to identify depression intensity from user-provided text. The system analyzes written content such as personal thoughts, journal entries, or social media text to detect emotional patterns and assess mental health risk levels. By applying natural language processing and machine learning techniques, the system provides an effective approach for early mental health awareness and support. The proposed system integrates text preprocessing, emotion analysis, and a trained machine learning model within a user-friendly web application to deliver real-time analysis results. In addition to prediction, the platform offers visual dashboards, emotion breakdowns, and personalized recommendations to help users understand their mental health condition. The system also supports result storage and advisor-level summaries, making it suitable for both individual use and guided mental health assessment. This project demonstrates how AI-based text analysis can provide a scalable, reliable, and accessible solution for mental health monitoring.

Keywords: Mental Health Analysis, Depression Detection, Natural Language Processing, Machine Learning, Emotion Analysis, Web Application.

Abstract: The Career Path Recommendation System is a web-based platform designed to assist students, job seekers, and professionals in identifying suitable career paths based on their interests, skills, academic background, and personality traits. Choosing the right career is often challenging due to the wide range of options, lack of guidance, and limited awareness about individual strengths and industry requirements. This system addresses these challenges by providing personalized, data-driven career recommendations that align with users’ profiles and preferences. The platform enables users to register, complete a structured assessment questionnaire, and receive career suggestions tailored to their skills, strengths, and aspirations. It incorporates algorithms to analyze user inputs, map them to relevant career domains, and provide actionable insights, including required skills, educational pathways, potential job roles,and growth opportunities. Administrators can manage career options, update assessment questions, and maintain the recommendation database. Developed using modern web technologies, the system offers an interactive frontend for a seamless user experience, a robust backend for processing assessments and generating recommendations, and secure data storage to protect user information. By combining technology and career guidance, the Career Path Recommendation System empowers users to make informed, confident, and strategic career decisions, improving employability and professional growth

Keywords: Career Path Recommendation System, Artificial Intelligence (AI), Machine Learning, Skill and Interest Analysis, Personalized Career Guidance.

Abstract: The rapid growth of online travel platforms has increased the complexity of trip planning, requiring users to manually search across multiple sources for destinations, weather conditions, accommodations, and transportation options. Traditional travel planning systems often lack personalization and real-time adaptability. This project presents an AI-based travel planner, Voyage-AI, which generates personalized travel itineraries using user preferences and real-time data sources. The system integrates Artificial Intelligence, location-based services, and weather APIs to provide intelligent travel recommendations. A web-based interface allows users to input travel details, generate itineraries, manage bookings, and visualize results efficiently. The proposed system offers a scalable, automated, and user-friendly solution for modern travel planning

Keywords: AI Travel Planner, Itinerary Generation, Machine Learning, Smart Tourism, Travel Recommendation System

Abstract: Web-based systems are frequently targeted by malicious users due to their public accessibility and widespread usage. In many cases, security incidents are detected only after damage has occurred, as conventional protection mechanisms focus primarily on access control rather than behavioral observation. This paper presents a Honeypot-Based Web Security Monitoring System supported by a web dashboard, implemented using PHP, HTML, CSS, and JavaScript. The system deploys intentionally deceptive web components within an isolated environment to attract unauthorized interactions and record them for analysis. All captured events are stored and presented through a structured dashboard that supports manual inspection. The solution emphasizes simplicity, transparency, and educational value, making it suitable for internal academic evaluation and small-scale deployments.

Keywords: Honeypot System, Web Security Monitoring, Attack Logging, PHP-Based Dashboard, Cybersecurity Analysis

Abstract: The Real-Time Chat Application is a web-based communication system developed to enable instant and reliable message exchange between users using modern full-stack technologies. The application leverages the MERN stack along with Socket.IO to provide real-time, bidirectional communication without page refresh or noticeable delay. Users can securely register, authenticate, and communicate through one-to-one messaging with live message delivery and online status tracking. The system uses Node.js and Express.js for backend processing, MongoDB for secure storage of user data and chat history, and React.js for building a responsive and interactive user interface. Socket.IO is integrated to manage real-time communication events efficiently, ensuring low latency and consistent message synchronization across active sessions. Additional features such as message persistence, timestamps, and session management enhance usability and reliability. The proposed system demonstrates how real-time communication frameworks combined with modern web technologies can deliver a scalable, efficient, and user-friendly messaging platform suitable for instant communication and collaborative environments.

Keywords: Real-Time Chat Application, MERN Stack, Socket.IO, WebSockets, Instant Messaging

Abstract: Cloud computing has rapidly evolved into the foundational infrastructure for modern digital services, enabling organizations to deploy applications with unprecedented scalability, elasticity, and cost efficiency. However, the dynamic and distributed nature of cloud-native environments introduces complex security challenges that traditional intrusion detection systems are ill-equipped to address. Conventional IDS solutions depend on static network boundaries, deep packet inspection, and predefined attack signatures, all of which are increasingly ineffective in cloud environments dominated by encrypted traffic, ephemeral workloads, and software-defined networking. This study presents an extensive analysis of cloud-native intrusion detection systems that utilize Virtual Private Cloud (VPC) Flow Logs in conjunction with ensemble learning techniques. VPC Flow Logs provide scalable, lightweight, and privacy-preserving network traffic metadata, making them suitable for large-scale cloud monitoring. Ensemble learning methods combine multiple machine learning classifiers to enhance detection accuracy, reduce false positives, and improve robustness against evolving cyber threats. This paper systematically reviews existing research, explores architectural designs, analyzes detection methodologies, evaluates benefits and limitations, and discusses future research directions. The study demonstrates that ensemble-based intrusion detection using VPC Flow Logs offers a practical and effective solution for securing modern cloud infrastructures.

Keywords: Cloud Security, Intrusion Detection System, VPC Flow Logs, Ensemble Learning, Machine Learning, Cloud-Native Architecture.

Abstract: Visually impaired people often face difficulties in understanding their surroundings, especially while moving in unfamiliar environments. Identifying nearby objects, obstacles, or people usually requires external help, which limits independence. With the availability of smartphones equipped with cameras and processing power, it is possible to use artificial intelligence to provide real-time assistance. This paper presents AI Eyes for Visually Impaired, an Android-based mobile application designed to assist visually impaired users through real-time object detection and voice feedback. The application uses the mobile camera to continuously capture the surrounding environment and applies an AI-based object detection model to recognize common objects. Based on the detected objects, the system identifies their direction and estimates their approximate distance, and then communicates this information through audio guidance using text-to-speech. The proposed system runs entirely on the mobile device without requiring internet connectivity, making it portable and practical for daily use. The results show that the application can effectively improve environmental awareness and support safer navigation for visually impaired users.

Keywords: Artificial Intelligence, Object Detection, Assistive Technology, Visually Impaired, Android Application, TensorFlow Lite, Real-Time Navigation, Text-to-Speech.

Abstract: Online apparel shopping has become increasingly popular, but customers still face difficulties in visualizing how clothes will look and fit on their bodies before purchasing. Static product images and size charts fail to provide a realistic understanding of garment appearance, often leading to dissatisfaction and high return rates. This limitation highlights the need for an interactive and personalized virtual try-on solution. This paper presents HerFit, a real-time virtual dressing system that allows users to try on garments virtually using live webcam input. The system detects the user’s body posture using computer vision techniques and overlays selected clothing items onto the user’s live video feed. By dynamically aligning garments with body landmarks, HerFit provides a realistic and interactive dressing experience without the need for physical trials.

Abstract: This paper presents an integrated credit card fraud classification approach based on ensemble learning and deep learning techniques to address the challenges of class imbalance and evolving fraud patterns in financial transaction data. Ensemble models such as Random Forest and Gradient Boosting are employed to enhance prediction reliability by combining multiple base classifiers, while deep learning architectures, including Deep Neural Networks, are utilized to capture complex and non-linear relationships within transaction features. Comprehensive data preprocessing and imbalance handling strategies are applied to improve model robustness. The proposed framework is evaluated using real-world credit card transaction datasets and assessed through standard performance metrics. Experimental results demonstrate that the hybrid ensemble–deep learning model outperforms traditional machine learning classifiers in terms of accuracy, precision, recall, F1-score, and area under the ROC curve. The findings confirm that the proposed approach provides an effective, scalable, and reliable solution for real-time credit card fraud detection in modern digital payment systems. The rapid growth of digital payment systems has significantly increased the volume of credit card transactions, making fraud detection a critical challenge for financial institutions. Credit card fraud classification aims to accurately distinguish fraudulent transactions from legitimate ones while minimizing false alarms. This task is particularly complex due to the highly imbalanced nature of transaction data, evolving fraud patterns, and the need for real-time decision-making. In this work, machine learning-based classification techniques are employed to analyze transaction behavior and identify potential fraud. Preprocessing steps such as data normalization, feature selection, and imbalance handling are applied to improve model performance. Multiple classifiers are trained and evaluated using standard performance metrics including precision, recall, F1-score, and area under the ROC curve. The experimental results demonstrate that intelligent classification models can effectively detect fraudulent activities with high accuracy and reduced false positives. The proposed approach enhances transaction security and supports financial organizations in mitigating monetary losses while ensuring a seamless experience for genuine customers.

Abstract: Personalized learning is important because every student learns at a different speed and has different goals. Many learners face difficulty while studying online due to a lack of guidance, unclear explanations, and no immediate feedback. To solve this problem, this project presents an LLM Powered Chatbot for Personalized Learning. The system works as a virtual tutor that interacts with learners through a web application. It allows users to enter a learner profile such as name, knowledge level, learning goal, and learning style. Based on these details, the chatbot generates customized responses that are easier to understand. The system also provides additional learning tools such as quiz generation and learning roadmap creation. Quizzes help learners test their knowledge, and roadmaps provide a structured plan for learning a topic step by step. The proposed system uses a locally hosted LLM (via Ollama) to generate answers, ensuring privacy and reducing dependency on cloud services. The results show that the chatbot improves the learning experience by giving clear explanations, supporting self-assessment, and helping learners follow a structured learning path.

Keywords: Large Language Model (LLM), Personalized Learning, Chatbot, Virtual Tutor, MCQ Generation, Learning Roadmap, Flask, Ollama, Education Technology

Abstract: Agriculture plays a vital role in the economic growth of many developing countries, yet farmers often face challenges related to improper crop selection, inefficient fertilizer usage, and delayed identification of plant diseases. Recent advancements in artificial intelligence offer promising solutions to address these issues through data-driven decision support systems. This paper presents a Smart Farming AI system that integrates machine learning and deep learning techniques to assist farmers in making informed agricultural decisions. The proposed system provides crop recommendations based on soil nutrient levels and environmental parameters, fertilizer suggestions using nutrient deficiency analysis, and plant disease detection through image-based analysis. The system is implemented as a web-based application using the Flask framework, ensuring accessibility and ease of use. Additionally, the application supports camera-based image capture and multilingual interaction to improve usability for farmers with limited technical knowledge. Experimental evaluation demonstrates that the system can generate accurate and explainable recommendations in a controlled environment. The proposed solution serves as an effective educational and decision-support platform for demonstrating the application of artificial intelligence in agriculture.

Keywords: Smart Farming, Machine Learning, Deep Learning, Crop Recommendation, Fertilizer Recommendation, Plant Disease Detection, Artificial Intelligence, Precision Agriculture.

Abstract: Web application security continues to be a critical concern as cyber attacks targeting online platforms grow in frequency and sophistication. Traditional Web Application Firewalls (WAFs) provide defense exclusively at the server level, leaving client-side vulnerabilities unaddressed and creating single points of failure. This paper presents a Dual-Layer Proxy-Based Web Application Firewall, a novel security framework that implements protection at both client and server layers through an integrated hybrid intelligence approach. The proposed system combines a browser-based extension with a backend proxy server to detect and block multiple attack categories including SQL Injection, Cross-Site Scripting (XSS), Path Traversal, Command Injection, and Server-Side Request Forgery (SSRF). Unlike conventional single-layer WAFs, this framework employs pattern-matching algorithms at the client side to intercept malicious requests before transmission, while the server layer performs deep packet inspection using advanced regex-based detection rules. The system features a comprehensive management dashboard with real-time attack analytics, domain protection management, and automated PDF report generation. Implemented using Python Flask, Chrome Manifest V3, SQLAlchemy ORM, and Chart.js visualizations, the framework achieves 99.8% attack detection accuracy with minimal performance overhead (less than 5ms client-side latency and approximately 20ms server-side processing time). Experimental validation through structured testing with 50+ attack payloads demonstrates the system's effectiveness in identifying and mitigating security threats while maintaining usability and transparency. This work highlights the significance of defense-in-depth strategies in modern web security and provides a scalable, open-source alternative to commercial WAF solutions.

Keywords: Web Application Firewall, Dual-Layer Security, SQL Injection Detection, Cross-Site Scripting Prevention, Browser Extension, Flask Framework, Cybersecurity.

Abstract: Gesture recognition for voice synthesis is an emerging assistive technology that enables communication through hand gestures by converting them into synthesized speech. This system is especially beneficial for individuals with speech or hearing impairments, providing them an alternative medium for interaction. With advancements in computer vision and deep learning, gesture-based interfaces have become more accurate and efficient. This paper presents a Gesture Recognition for Voice Synthesis System that uses computer vision and deep learning techniques to recognize predefined hand gestures in real time and convert them into corresponding voice outputs. A Convolutional Neural Network (CNN) is employed for gesture classification after preprocessing steps such as image resizing, background normalization, and feature extraction. Once a gesture is recognized, a text-to-speech module generates an appropriate voice output. The system is implemented as a real-time application using a camera interface, allowing users to perform gestures naturally. Experimental evaluation shows high recognition accuracy and low response latency, demonstrating the effectiveness of the proposed system for assistive communication and human–computer interaction applications.

Keywords: Gesture Recognition, Voice Synthesis, Computer Vision, Deep Learning, CNN, Assistive Technology

Abstract: The rapid growth of computer science education and competitive programming has increased the demand for efficient online platforms that support coding practice, assessment, and skill evaluation. Traditional classroom-based learning and manual evaluation methods often lack scalability, instant feedback, and real-time performance analysis. To address these challenges, this paper presents CodeHub, a full-stack online coding practice and evaluation platform designed to help students enhance their programming and problem-solving skills. CodeHub provides a centralized environment where users can practice Data Structures and Algorithms (DSA), solve coding problems, participate in contests, and receive instant code evaluation through an integrated online judge system. The platform supports secure user authentication, role-based access for students and administrators, and automated code execution using APIs such as Judge0. Built using modern web technologies including React, Node.js, PostgreSQL, and RESTful APIs, CodeHub ensures scalability, reliability, and a user-friendly experience. Experimental usage shows that CodeHub improves learning efficiency by providing real-time feedback, standardized evaluation, and continuous skill assessment, making it a valuable tool for modern programming education.

Keywords: Online Coding Platform, CodeHub, Data Structures and Algorithms, Code Evaluation, Judge0, Full-Stack Development, Programming Education

Abstract: Personal safety has become a major concern in today’s fast-paced and technology-driven society. With the increasing need for quick and reliable emergency response systems, there is a demand for applications that can provide immediate assistance during critical situations. The Threat Guard project is an AI-based personal safety web application designed to enhance individual security by enabling instant emergency alert generation and real-time user interaction. Threat Guard is developed as a client-side web application using modern frontend technologies such as React, TypeScript, and Vite. The system allows users to trigger emergency alerts through a simple interface, activate audio and visual notifications, capture real-time evidence, and share live location details with registered emergency contacts. The application follows a component-based architecture, ensuring modularity, scalability, and ease of maintenance. The system dynamically responds to user-triggered events without requiring full page reloads, providing fast response times and improved user experience. Emphasis is placed on simplicity, reliability, and performance to ensure effective operation during emergency scenarios. The project successfully demonstrates the practical implementation of an alert-based personal safety system and serves as a strong foundation for future enhancements such as backend integration, advanced threat detection, and mobile application support.

Abstract: Efficient management of shared institutional resources such as auditoriums is a critical challenge in educational organizations. Manual and semi-digital booking methods often result in scheduling conflicts, underutilization, and lack of transparency. This paper presents a Web-Based Auditorium Utilization and Alert System designed to automate auditorium booking, approval workflows, conflict detection, and user notification. The proposed system adopts a role-based access control mechanism and a centralized database to ensure secure and transparent operations. By automating approvals and providing real-time alerts, the system significantly reduces administrative workload and improves resource utilization efficiency. Experimental deployment in an institutional environment demonstrates improved scheduling accuracy, reduced conflicts, and enhanced user satisfaction.

Keywords: Auditorium Management, Resource Scheduling, Web Application, Alert System, Institutional Automation.

Abstract: This project presents a Face Recognition Based Attendance System designed to automate and streamline the attendance process in educational institutions. The system replaces traditional manual and biometric methods with a contactless, accurate, and time-efficient solution using face recognition technology. Students are registered with their academic details and facial data, which are used to train a recognition model for real-time attendance marking. The system consists of Admin, Student, and Attendance modules to manage registrations, monitor attendance, and generate reports. By minimizing human intervention and errors, the proposed system improves reliability, security, and efficiency in attendance management.

Keywords: Face Recognition, Attendance Management, Computer Vision, Machine Learning, Biometric Authentication, Automated Attendance, OpenCV, Python Programming, Real-time Face Detection, Educational Institutions

Abstract: Communication barriers between hearing-impaired and hearing individuals remain a significant challenge in everyday interactions. While spoken and written languages are widely supported by modern technologies, sign language communication still lacks accessible and real-time translation solutions. This project presents a Multilingual Speech-to-Sign Language Translator with Avatar, designed to bridge this communication gap using artificial intelligence and human–computer interaction techniques. The proposed system accepts user input in the form of speech or text, converts it into a target language using multilingual translation models, and represents the translated content through a 3D animated sign language avatar. In addition, the system integrates real-time hand gesture recognition using computer vision techniques to identify basic sign gestures and map them to corresponding textual meanings. This bidirectional interaction enables both hearing and hearing-impaired users to communicate more naturally. The system architecture combines speech recognition, language translation, text-to-speech synthesis, gesture detection, and avatar animation into a unified web-based platform. By processing inputs locally and rendering sign outputs visually, the system ensures low latency and improved user experience. Experimental evaluation demonstrates accurate speech recognition, smooth avatar animation, and effective translation across multiple languages. The proposed solution offers an affordable and scalable assistive communication tool that can be deployed in educational institutions, public service centers, and social interaction platforms. By enhancing accessibility and inclusivity, this work contributes toward improving digital communication for the hearing-impaired community while supporting multilingual interaction in real time.

Keywords: Speech-to-Sign Translation, Sign Language Avatar, Gesture Recognition, Multilingual Translation, Assistive Technology, Human-Computer Interaction

Abstract: Preparing for academic assessments effectively remains a critical challenge for students and faculty alike, with many relying on traditional manual methods that lack personalization, scalability, and real-time feedback mechanisms. Current evaluation tools provide limited domain-specific guidance, lack semantic understanding of descriptive answers, and fail to capture the nuanced pedagogical criteria defined by frameworks like Bloom’s Taxonomy. The absence of adaptive, AI-driven assessment systems leaves students underprepared for complex, curriculum-aligned evaluations tailored to their specific subjects and cognitive levels. To address these limitations, the EduRAG System integrates Generative AI, Retrieval-Augmented Generation (RAG), and Computer Vision to deliver personalized, interactive assessment at scale. The system leverages advanced NLP and transformer-based models to generate contextually relevant technical questions based on syllabus content and cognitive difficulty levels provided by users. Real-time Optical Character Recognition (OCR) captures handwritten student responses, while AI-powered evaluation mechanisms assess semantic accuracy and conceptual depth against industry-standard "Ground Truth" extracted from the curriculum. The application provides instant, detailed feedback including quantitative scoring, improvement suggestions, and performance analytics across multiple evaluation attempts. Through a user-centric web platform, faculty access role-specific question generation banks and students receive AI-generated recommendations for skill enhancement. By combining adaptive question generation with semantic answer analysis, the proposed solution significantly improves academic performance while democratizing access to high-quality pedagogical tools.

Keywords: Retrieval-Augmented Generation, Semantic Vectorization, Bloom’s Taxonomy, Automated Assessment.

Abstract: This paper presents an AI-powered Real Estate Management System with Intelligent Consultant and Sales Agent designed to address information asymmetry, fragmented data sources, and inefficient property discovery in the Indian real estate sector. The system integrates a centralized MySQL database containing 382+ verified property listings across Bangalore with an AI-powered chatbot that analyzes user queries and requirements to provide real-time property recommendations, instant customer support, and expert consultation. By leveraging Cohere Chat API within a PHP-based web application, the system delivers role-based AI responses that adapt dynamically based on user intent, functioning as a Sales Agent, Investment Advisor, Legal Advisor, or General Assistant. The platform includes property management with CRUD operations, customer relationship management, automated approval workflows, dynamic search capabilities, and real-time synchronization between the database and AI knowledge base. This approach demonstrates how artificial intelligence and web technologies can offer an accessible, cost-effective solution for property management and intelligent property discovery, transforming traditional real estate operations through PropTech innovation and promoting data-driven decision-making.

Keywords: Real Estate Management System, Artificial Intelligence, Property Recommendations, PropTech, Web Application, Intelligent Chatbot, Customer Engagement.

Abstract: Lawvia AI is an intelligent legal assistance and document analysis platform designed to improve legal awareness and accessibility. Users such as students, employees, and tenants often face legal issues including harassment, cybercrime, and rental disputes without adequate guidance. The system accepts text or voice input and applies Natural Language Processing (NLP) to identify the legal context. A BERT-based transformer model embedded in the chatbot performs legal intent and domain classification, enabling simplified explanations of relevant laws and recommended actions. Lawvia AI includes a Complaint Draft & Action (DocuAction) module for generating structured complaint letters with options to download or forward them to external authorities, and a Legal Support (LegalAid+) module that provides access to nearby legal aid resources. The platform also supports document risk analysis, where uploaded legal documents are processed using Optical Character Recognition (OCR) and NLP-based keyword mapping to detect risky clauses and generate simplified summaries. Additionally, the ESpace authority-routing module assigns complaints to appropriate internal authorities based on severity. The system is implemented using ASP.NET with C#, integrates Python-based NLP components, BERT models, LLM APIs, and SQL Server, offering a modular and socially impactful legal assistance solution.

Keywords: Legal Assistance, Document Risk Analysis, Natural Language Processing (NLP), Machine Learning (ML), Complaint Letter Generation, Authority Routing.

Abstract: The rapid expansion of digital communication platforms has increased the spread of harmful and offensive content across text, images, audio, videos, and online comments. Manual moderation methods are inefficient and difficult to scale for large volumes of multimodal data. This project presents a Multimodal Harmful Content Classifier with Streamlit to automatically identify and classify harmful content. The system processes multiple input formats and converts non-textual data into textual form using OCR and Speech-to-Text techniques. Machine learning and natural language processing methods are applied to analyze extracted content. The classifier categorizes input as Safe, Offensive, or Harmful with a confidence score. A Streamlit-based interface provides real-time analysis and result visualization. The proposed system improves moderation accuracy and supports safer digital communication.

Abstract: The increasing levels of food waste alongside the persistent issue of hunger highlight the urgent need for efficient food redistribution mechanisms. However, conventional food donation processes often rely on manual coordination, leading to delays, miscommunication, and significant food spoilage before it reaches those in need. This paper presents a Web-Based Food Donation and Redistribution System designed to bridge the gap between food donors and non-governmental organizations (NGOs) through a centralized digital platform. The proposed system enables donors such as restaurants, event organizers, and households to register surplus food in real time, while nearby NGOs can view, accept, and coordinate collection efficiently. Location-based tracking and automated notification features ensure timely communication and reduce response delays. The system leverages modern web technologies and database management to maintain secure records of donations, users, and distribution activities. An administrative module oversees verification and system monitoring to ensure reliability and transparency. The system is evaluated using simulated operational scenarios, demonstrating improvements in coordination speed, reduction in food wastage, and increased distribution efficiency compared to traditional manual donation methods. The proposed platform offers a scalable and practical solution for addressing food waste while supporting hunger relief initiatives.

Keywords: Food Donation, Food Waste Management, NGO Coordination, Web Application, Real-Time Distribution, Smart Donation System

Abstract: The growing demand for digital learning platforms has significantly influenced the evolution of language learning methodologies. However, many existing platforms for regional languages such as Kannada lack structured content, interactivity, and effective assessment mechanisms. Most available tools rely on static vocabulary lists and traditional teaching approaches, which reduce learner engagement and hinder systematic language acquisition. This paper presents Kannada Kaliyiri, a web-based interactive platform designed to provide a structured and engaging environment for Kannada language learning. The system organizes learning content into categorized modules including animals, fruits, vegetables, greetings, grammar elements, vowels (Swaragalu), and consonants (Vyanjanagalu). Each learning unit includes Kannada script, English meaning, and transliteration to facilitate comprehension and pronunciation. An integrated quiz module dynamically generates questions from multiple categories and provides instant feedback to learners. The application is developed using modern web technologies such as React.js, JavaScript, HTML5, CSS3, and Tailwind CSS. The modular architecture ensures scalability, maintainability, and extensibility of the system. Experimental evaluation demonstrates that Kannada Kaliyiri significantly improves learner engagement and understanding compared to traditional static learning methods. The platform highlights the potential of interactive web-based systems in promoting regional language education and can be extended with advanced features such as audio pronunciation, adaptive learning mechanisms, and personalized progress tracking.

Abstract: Agriculture plays a crucial role in the economic growth of many developing countries, where crop productivity is often threatened by plant diseases. Early and accurate identification of plant diseases is essential to minimize yield loss and ensure sustainable agricultural practices. However, traditional disease detection methods rely heavily on manual inspection and expert knowledge, which are time-consuming, subjective, and not easily accessible to farmers in rural areas. Recent advancements in artificial intelligence, particularly deep learning, offer effective solutions for automated plant disease diagnosis. This paper presents a Plant Disease Detection System based on deep learning techniques for accurate and automated identification of plant diseases from leaf images. The proposed system employs a Convolutional Neural Network (CNN) trained on the Plant Village dataset to classify plant leaves into healthy or diseased categories. The system is implemented as a web-based application using the Flask framework, allowing users to upload plant leaf images and obtain instant disease predictions through a simple and user-friendly interface. Image preprocessing and model inference are handled efficiently to ensure reliable performance. Experimental evaluation demonstrates that the proposed system can accurately identify common plant diseases in a controlled environment, enabling early disease detection and timely preventive measures. The developed solution serves as an effective educational and decision-support platform, highlighting the practical application of artificial intelligence and computer vision in modern agriculture.

Keywords: Plant Disease Detection, Deep Learning, Convolutional Neural Network, Computer Vision, PyTorch, Flask, Agriculture, Image Classification

Abstract: CodeAlong is a web-based collaborative code editor designed to support real-time programming and teamwork among developers, students, and researchers. The system enables multiple users to write, edit, and review source code simultaneously within a shared workspace. By integrating live synchronization, communication tools, and execution support, CodeAlong reduces delays caused by traditional file sharing and version conflicts. The platform promotes effective collaboration by allowing users to monitor changes instantly, exchange ideas, and resolve errors collectively. In addition, it provides user authentication, role management, and secure data handling to ensure reliability and privacy. The proposed system aims to enhance productivity, improve learning experiences, and support distributed software development. Experimental evaluation indicates that CodeAlong improves coordination and reduces development time when compared to conventional standalone editors. Therefore, the platform serves as a practical solution for modern collaborative programming and research activities.

Keywords: Collaborative Programming, Real-Time Code Editor, Web-Based Development, Multi-User Editing, Code Sharing, Software Collaboration, Online IDE, Version Control, Distributed Development, Learning Platforms

Abstract: Chatbots have become an essential tool for providing automated and efficient user interaction in various application domains. This project focuses on the design and development of an intelligent chatbot system that utilizes Natural Language Processing and machine learning techniques to understand user queries and generate appropriate responses. The system performs intent recognition, input validation, and error handling to ensure reliable communication. Functional testing and performance optimization were carried out to improve response accuracy and reduce response time. The results demonstrate that the proposed chatbot system offers accurate responses, stable performance, and an improved user experience. The system can be further enhanced by incorporating advanced learning models, multilingual support, and voice-based interaction.

Abstract: Public bus transportation continues to be the primary mode of travel for millions of commuters in major Indian metropolitan cities. Although several digital bus tracking applications are available, most existing systems suffer from inaccurate Estimated Time of Arrival (ETA) predictions, limited awareness of passenger crowd levels, slow data update rates, and a lack of intelligent predictive features. These shortcomings often lead to longer waiting times, overcrowded buses, and an overall unsatisfactory commuting experience. To address these challenges, this paper proposes an AI-driven predictive platform for bus mobility and real-time human flow analysis. The platform integrates live GPS tracking with advanced machine learning models to deliver accurate and reliable transit information. Long Short-Term Memory (LSTM) networks are used to predict bus arrival times, while Random Forest algorithms estimate passenger crowd levels. In addition, the system incorporates real-time WebSocket communication, weather-aware routing, and user travel pattern learning through K-Means clustering. Developed as a Progressive Web Application (PWA), the platform supports multi-city scalability, offline access, and real-time notifications. Experimental results demonstrate an ETA prediction accuracy of up to 85% and crowd classification accuracy of 78%, significantly outperforming existing public transport applications. Overall, the proposed solution improves commuter decision-making, reduces travel uncertainty, and contributes to the development of smarter and more efficient urban mobility systems.

Keywords: Bus Tracking, ETA Prediction, Human Flow Analysis, LSTM, Random Forest, Intelligent Transportation Systems.

Abstract: This study investigates the asymptotic optimal control of a data transmission queue operating under heavy traffic conditions with an imperfect channel. This model considers a single-server N-policy queue where packets arrive according to a Poisson process and transmissions are subject to channel failures, leading to retransmissions that rejoin the queue. The server remains inactive until the queue reaches a threshold N, after which it continues service until the system empties. Steady-state balance equations are developed to obtain explicit probability distributions for OFF and ON states, and heavy traffic scaling is used to derive asymptotic expressions for queue length, server utilization, and cost. The analysis establishes a reduced cost function capturing the trade-off between holding and activation costs and shows that the optimal activation threshold follows a classical square-root heavy traffic law. Numerical illustrations and simulations compare exact steady-state optimization with heavy traffic approximations, highlighting the conservatism and near optimal performance of asymptotic policies as system utilization approaches saturation.

Keywords: Data transmission queue, N-policy, imperfect channel, retransmission, heavy traffic, asymptotic optimal control