Build a RAG Document Q&A Chatbot (OpenAI + Python)

I'm looking for an experienced AI developer to build a small, working RAG (Retrieval-Augmented Generation) chatbot that answers questions over a set of PDF documents. What I need: - Ingest ~10–20 PDF documents (text-based, no OCR needed) into a vector store (Chroma, Pinecone free tier, or similar - your recommendation welcome) - A simple Q&A pipeline using the OpenAI API that retrieves relevant chunks and answers questions grounded in the documents only - no hallucinated answers; if the answer isn't in the docs, the bot should say so - Source citations with each answer (which document/page the answer came from) - A minimal chat interface to test it - a simple React/Next.js page or even Streamlit is fine; this is an internal prototype, not a polished product - Clean, documented Python code I can run locally (README with setup instructions, .env for API keys) Scope notes: - This is a prototype/MVP - no auth, no deployment, no multi-user support needed - I'll provide the documents and my own OpenAI API key Ideal freelancer: - Strong experience with RAG pipelines and the OpenAI API - Python backend skills (Flask/FastAPI or scripting is fine) - Cares about answer accuracy and evaluation, not just wiring APIs together - Bonus: React/Next.js experience in case we expand this later To apply: Briefly describe a RAG system you've built, what vector store you'd use here and why, and how you'd prevent the bot from making up answers. Deliverables: Source code (GitHub repo or zip), README