Build a RAG Chatbot Over Our PDFs (Python, LangChain, OpenAI)

We’re a small US-based company with about 80 internal PDFs (product guides, SOPs, FAQs). Our support team keeps answering the same questions, and we want an AI chatbot that answers accurately from our own documents, not generic ChatGPT answers. We need an experienced RAG developer who can deliver a working, production-ready version within 4–5 days. What we need: 1: Ingest our PDFs with clean chunking and metadata 2: Vector database setup (Pinecone, Qdrant, Chroma, or pgvector; open to your recommendation) 3: RAG pipeline using LangChain or LlamaIndex with the OpenAI API Answers must include source citations (document name and page) 4: If the answer isn’t in our documents, the bot should say so instead of guessing FastAPI backend with a /chat endpoint 5: A simple chat UI (Streamlit or a basic React page is fine) Quick evaluation on 15–20 sample questions we provide (RAGAS or similar is a plus) Docker setup and a clear README Deliverables: Working chatbot, deployed or ready to deploy Full source code in our GitHub repo (we own all code) Short evaluation summary on answer quality A Loom walkthrough video Budget: $500 fixed price Timeline: 4–5 days Ideal freelancer: Has built RAG systems before (please share examples or a GitHub link) Strong with Python, LangChain/LlamaIndex, FastAPI, and vector databases Cares about accuracy and preventing hallucinations Can share a daily progress update To apply, please answer: Briefly describe a RAG project you’ve built and how you improved answer accuracy. Which vector database would you choose for this and why? Can you deliver within 5 days? Share a rough day-by-day plan. Please start your proposal with the word “Grounded” so we know you read the full post.