Series · 8 of 8 parts · Intermediate
RAG, Properly
Why retrieval-augmented generation goes wrong in production, and the specific decisions that keep answers grounded in something real.
What you’ll understand
By the end you understand why RAG fails and can build a pipeline that cites its sources and stays honest.
Episodes
01Why RAG Exists: The Context ProblemModels can't know your data. RAG is the architecture that fixes it, and its failure modes start here.10 min→02Chunking: The Decision Everything Downstream InheritsSizes, overlap, and semantic boundaries. Get this wrong and nothing after it can save you.20 min→03Embeddings and Vector Search, DemystifiedCosine similarity, dimensions, and what 'semantic' actually buys you.12 min→04Vector Databases: What Actually Matters When ChoosingIndexes, filters, scale, and cost across the popular options.11 min→05Hybrid Search and Re-Ranking: The Cheapest Quality WinKeywords catch what vectors miss, and re-rankers fix the order. Most teams skip both.12 min→06Grounding and Citations: Answers You Can CheckSource quotes, confidence gates, and the honest 'I don't know'.10 min→07Evaluating RAG: Retrieval Metrics That Predict Answer QualityRecall@k, faithfulness, and the eval harness that catches regressions.9 min→08Agentic RAG: When the Retriever Starts ThinkingQuery rewriting, multi-step retrieval, and agents that decide what to look up.13 min→