Everyone can build a RAG pipeline that works on a demo. Embed some docs, run a cosine similarity search, stuff the results into a prompt - done, right? Then you deploy it to production and discover your retrieval is surfacing irrelevant documents, missing obvious matches, and hallucinating sources your users trusted.

In this session, Spencer will share from-the-trenches case studies from production RAG systems handling real business data at scale. We'll go over why naive embedding search fails, how strategies like hybrid retrieval (full-text + semantic) dramatically improves relevance, and the chunking and re-ranking strategies that actually matter. You'll also learn how to measure whether your retrieval is actually working, because if you're not measuring, you're guessing.

This isn't another "RAG 101" talk - it's the talk you need after your first RAG pipeline disappoints you.