Built a production retrieval-augmented (RAG) knowledge base over a large, multi-repository engineering codebase and document corpus, so an LLM could answer grounded questions instead of hallucinating across a sprawling polyglot codebase.
Context and stakes
Engineering knowledge was scattered across a large multi-repository ecosystem and hundreds of pull requests and design documents. To make any of it useful to an LLM, for search, grounding, or content generation, it had to be ingested, chunked, embedded, and retrievable with enough precision that the answers were trustworthy rather than hallucinated.
Problem
Naive approaches break at this scale. Fixed-size chunking shreds code blocks and headers, a single embedding model forces one compromise on cost, latency, and control, and without metadata you cannot scope retrieval to the right repository, content type, or domain. The corpus also kept growing, so ingestion had to be repeatable, not a one-off script.