AI-POWERED JEE CHATBOT: A RAG-BASED ASSISTANT FOR SUBJECT - SPECIFIC DOUBT SOLVING

This project aims to develop an AI-powered chatbot to assist JEE aspirants with doubts in Physics, Chemistry, and Mathematics. It uses Retrieval-Augmented Generation (RAG) with Meta’s Llama 3.2 to provide accurate, step-by-step solutions by retrieving relevant information from JEE
textbooks, past papers, and solved examples.Data is extracted and cleaned using PyPDF2 and Unstructured.io, then converted into embeddings with Hugging Face’s sentence transformers and stored in ChromaDB for efficient retrieval. The RAG pipeline, implemented with LangChain, integrates the vector database with Llama 3.2 for context-aware responses. LoRA fine-tuning enhances accuracy on JEE-specific Q&A pairs. The backend, built with FastAPI, offering a robust
and scalable API for handling user queries. The frontend, developed using Streamlit (for MVP) or React (for production), provides a user-friendly interface where students can input doubts via text or voice. Advanced features like LaTeX rendering for equations, progress tracking, and mock test generation are incorporated to enhance usability. ensures scalable query handling, while the frontend, developed with Streamlit (MVP) or React (production), offers a user-friendly interface with LaTeX support for equations, progress tracking, and mock test generation. Deployment utilizes Modal for serverless LLM hosting, Vercel for the frontend, and Fly.io for the backend. Redis caching improves response times. This project demonstrates AI/ML, cloud deployment, and fullstack
development expertise, providing a scalable and practical solution to help JEE aspirants.