Description
We’re looking for a developer to build a small but production-quality RAG (Retrieval-Augmented Generation) chatbot connected to our internal knowledge base. What we need: We have a set of internal documents (playbooks, templates, reference materials) that we currently search through manually. We want a chatbot that can answer questions based on this content, with accurate retrieval, not generic AI responses. Scope: * Ingest and chunk our documents into a vector store (Pinecone or pgvector, your recommendation welcome) * Build a RAG pipeline: query → retrieval → grounded answer with source reference * Simple chat interface, a lightweight web-based chat UI is preferred (your recommended approach welcome, as long as it’s self-contained and easy to use) * Sensible fallback when the answer isn’t in the knowledge base, no hallucinated answers Tech preferences: LangChain or LangGraph, Pinecone or pgvector, Python or Node.js, open to your recommended stack as long as it’s production-grade, not a weekend hack. Timeline: 5–7 days Budget: $500 fixed price (open to milestone split if you prefer, e.g. setup/pipeline, then interface + polish) What we’re looking for in you: * Real experience with RAG pipelines, not just calling an LLM API directly * Can explain your retrieval approach (chunking strategy, embedding choice) before starting, not just “I’ll figure it out” * Clean handoff, documented code, not a black box Please include a short note on how you’d approach chunking/retrieval for a