Description
Summary Design and build a custom Retrieval-Augmented Generation (RAG) AI assistant that answers questions instantly and accurately from a defined document set or knowledge base. The system ingests the provided content (PDFs, docs, or website pages), converts it into vector embeddings, and uses an LLM with semantic search to answer in natural language — grounded in the source material with citations back to it. Built with Python, LLMs (LLaMA/GPT-family), a ChromaDB vector database, and a RAG pipeline. Deliverables include a working deployed assistant, clean source code, and a short handover guide. Details / Scope Ingest the client's documents / knowledge base into a vector store RAG pipeline: embeddings + semantic search + LLM answering Grounded answers with source citations Deployment: Hugging Face Space demo — or a lightweight API + embeddable JS widget for website integration Source code + brief setup/handover guide One round of testing and minor fixes Single milestone, fixed price