I’m building a retrieval-augmented generation (RAG) pipeline and need an experienced engineer who can take it from concept to a working proof-of-value. Scope • Connect the generation model to multiple retrieval sources—my internal databases, selected online articles, and live API data. • Handle data that sits in blob storage and a No-SQL database; you’ll decide the best way to index and chunk both semi-structured and unstructured content. • Orchestrate the workflow so a single prompt triggers retrieval, relevance ranking, and answer synthesis. • Ship a functional prototype with clear read-me style documentation outlining how to extend the data connectors and swap in different language models. Tech I expect you to be comfortable with • Python and popular LLM frameworks (LangChain, LlamaIndex or similar) • Vector stores (e.g., FAISS, Pinecone, or an equivalent you recommend) • Docker or similar containerisation for easy hand-off • Basic CI for repeatable local builds Deliverables 1. Well-commented codebase for the RAG pipeline 2. One-click launch instructions (Docker compose or similar) 3. Short report explaining architecture choices and next-step recommendations I’m available for quick feedback loops and will test the prototype against real internal queries as part of the acceptance criteria.