Python
An end-to-end privacy-preserving vision pipeline that accurately segments and identifies bystanders in images and replaces them using advanced inpainting and generation techniques, achieving robust anonymization while maintaining high aesthetic and informational quality.
Python
This study compares Q-Learning, DQN, and PPO for autonomous navigation in a grid-based maze with static and dynamic obstacles, evaluating learning efficiency, stability, and adaptability. The results reveal trade-offs between classical and deep RL methods, offering insights into selecting effective algorithms for real-world robotic navigation.
Python
This project investigates time-series forecasting of key air pollutants across major Indian cities using historical air quality data and multiple predictive models. Comparative evaluation and fuzzy health risk analysis identify effective approaches for accurate air quality prediction, supporting informed environmental monitoring and management.
Python
This GUI-based tool allows you to capture synchronized RGB, depth, and segmentation mask images from an Intel RealSense D435i camera. It also automatically generates annotations in YOLO format, stores point clouds, and logs camera intrinsics and scene metadata.
Python
Despite having no negative samples and a very small number of positive samples, this project attempts to use machine learning to classify images of the fungus M. slooffiae. Our goal is to develop a system that can determine whether a new image is an outlier or a member of the M. slooffiae class by utilizing data augmentation and one-class classification techniques, specifically One-Class SVM, Isolation Forest, and Autoencoders. The project shows how to adapt contemporary machine learning techniques for biological datasets that are uncommon or underrepresented.
Next.js, Express, Prisma, PostgreSQL, Supabase, Mistral-8x7B-Instruct-v0.1
Designed and developed TrustGuard AI, a full-stack AI-powered platform selected for the Amazon prototype submission round, enabling real-time detection of fake reviews, seller fraud, and counterfeit activity using multimodal data sources
Django, PostgreSQL, HTML, CSS, JavaScript, Bootstrap
Developed a full-stack, role-based facility booking system for campus-wide resource management, supporting real-time availability, structured approval workflows, and conflict resolution for classrooms, seminar halls, and AV equipment. Architected the backend using Django with ORM-based models and PostgreSQL, integrated Gmail SMTP for automated email notifications, and implemented a responsive UI using Bootstrap and vanilla JavaScript to ensure seamless user experience across devices.
React, sql.js
A lightweight, in-browser SQL database that lets you upload CSV files and run SQL queries on them instantly. This app runs entirely in your browser, requiring no server or backend.
Next, React, MongoDB, Mongoose, Express.js, Node.js, Tailwind, JWT
Designed an intuitive platform for organizing events like weddings and parties, featuring real-time vendor discovery, payment integration, and scheduling tools.
HTML, CSS, JavaScript, PHP, MySQL
Designed and implemented a comprehensive relational database schema for a local bus management system, optimizing information storage and retrieval across multiple entities.