The Innovation Herald
Github Coding Assistant
Developed an agent-based developer assistant that automates code reviews, bug detection, and patch generation directly from GitHub repositories. Using static analysis tools like pylint and LLM-powered reasoning, the agent identifies issues, suggests fixes, and can create pull requests via the GitHub API. ReAct-style control flow is implemented through LangChain, with a lightweight FastAPI backend and a React frontend. Designed to be modular, fast to deploy on free-tier services, and extensible for deeper code intelligence features in the future.
Autonomous Procurement Agent
Built an autonomous procurement agent that negotiates with suppliers through email and chat, handling everything from quote collection to database updates. Focused on multi-step reasoning using LangChain, with embedded quote comparison via sentence-transformers and FAISS, and policy checks through REST API calls. Integrated stateful memory using Redis and managed backend operations with FastAPI and PostgreSQL. The project is structured to reflect enterprise-grade agent workflows and is designed for modularity, extensibility, and open-source collaboration.
CRITSO: The official website of CRITSO
I created a multi-page website with transition animations between pages using framer-motion and implemented responsiveness using breakpoints and maintained overall quality of code. I experimented with various hover effects creating professional components like carousels, sliders, cards, sidebars, etc.
Chatbot for The Justice Department's website
I developed a custom Knowledge Graph for implementing a vector embeddings, then implemented QLoRA SFT method to train the Llama3.1 8B model using Unsloth. I Setup a RAG to work with a KB full of website scraped data and deployed the model via a Flask API in a NGROK tunnel on Google Collab.
Indian Sign Language AI
An attempt to create a CNN on Indian hand signs to grasp letter understanding from live feeds using OpenCV. My introduction to Neural Networks and Deep Learning, this project focuses on recognizing and classifying Indian Sign Language (ISL) alphabets in real-time. The goal is to enhance accessibility and communication for the hearing-impaired community while deepening my understanding of deep learning frameworks.