Projects

Project 01: Touchless Attendance system

Collaborated with the Tata Institute of Fundamental Research, to develop an application harnessing the power of face and voice recognition to mark attendance; achieved a 87% cost reduction by leveraging a cost-effective ESP microcontroller.
Implemented face recognition using OpenCV with 99.38% accuracy and added a 95.5% accurate layer of RNN for Voice classification; constructed distinct PyQT dashboards providing analytics from real-time MySQL database.
Authored and published a paper on the same in International Journal IEEE in 2023 International Conference on Artificial Intelligence and Applications (ICAIA) conducted in April 2023. DOI: 10169372

Led development of a responsive stock search web application, leveraging Angular and Node.js, hosted on Google App Engine, engineering full-stack solutions optimized for cloud deployment and scalable performance
Implemented advanced features such as real-time stock data visualization, recommendations, user-managed watchlists, and portfolios, using AJAX and JSON for seamless data exchange, enhancing user engagement and decision-making capabilities
Built an analogous iOS application in Swift delivering seamless cross-platform user experience, focusing on responsive design

Devised a ReactJS social media web application for ISA-VESIT to connect students with alumni, to serve as a hub for engaging discussions on latest topics, fostering relationships for academia and professional growth.
Created multiple APIs using Django and crafted the application's authorization module using JSON Web tokens

Programmed a Flutter application to inform users about nearest fuel station and arrange them in ascending order of real-time vehicular density, addressing significant waiting time inconveniences
Employed YOLOv5 vehicle detection algorithm on a custom-built dataset to accurately assess real-time traffic congestion, achieving an F1 score of 0.94. Developed PHP backend portal to display analytics, enabling informed business decisions
Published a paper on the same in 2022 IEEE Region 10 Symposium (TENSYMP), conducted in July 2022. DOI: 9864463

Developed a Retrieval-Augmented Generation system, using ChatOpenAI Large Language Model for query optimization, enhancing response precision and relevancy, innovatively combining parametric and non-parametric memories.
Led DuckDuckGo API integration for unbiased information retrieval, shifting from Wikipedia to improve response diversity and factual accuracy tokens.