Designed, developed, and tested a full RAG system with router chains across SQL, website, and PDF knowledge bases.
Knowledge lived in multiple formats. Users needed one chatbot that could retrieve the right context and answer accurately without manual switching between sources.
Implemented a router chain that directs queries to appropriate RAG pipelines over SQL, website content (Selenium-assisted), and PDF documents — with FastAPI services and a Streamlit interface.
Delivered an end-to-end LLM application demonstrating multi-source retrieval, routing, and practical deployment tooling.