Built a retrieval-augmented generation (RAG) tool that lets developers ask natural-language questions about any codebase and get accurate, context-aware answers. Designed and built the full backend using FastAPI, with FAISS for vector storage and all-MiniLM-L6-v2 embeddings for semantic search. Integrated Gemini 2.5 Flash for response generation, PostgreSQL/Supabase for data persistence, and GitHub OAuth for secure repo access. Also built the React/Vite frontend and resolved a critical authentication bug involving JWT cookie handling. A working prototype demonstrating practical LLM application design — from embedding pipelines to secure auth flows.