Retrieval-Augmented Generation (RAG) · TRACK 04/05
Advanced RAG Architectures
Agentic RAG, GraphRAG, self-correcting pipelines, and retrieval beyond plain text.
// THE TRACK
01 · START HEREAgentic RAGYou'll understand how agentic RAG turns retrieval into a tool the LLM calls iteratively, and when that beats a fixed pipeline.INTERMEDIATEHybrid RAGYou'll understand how running BM25 and vector search in parallel — then fusing their results — closes the gaps that either method leaves on its own.INTERMEDIATEGraphRAGYou'll understand how GraphRAG builds and queries a knowledge graph from documents, and which question types it answers better than vector RAG.INTERMEDIATECitations & GroundingYou'll know the practical techniques for making RAG answers cite their sources, from inline markers to post-hoc verification.ADVANCEDParent Document RetrievalYou'll understand the small-to-big pattern: match on tiny chunks for precision, then hand the LLM the larger parent for context.INTERMEDIATEMultimodal RAGYou'll learn how multimodal RAG indexes and retrieves images, charts, and tables alongside text so answers use the whole document.INTERMEDIATERAG Over Structured DataYou'll learn why embedding rows rarely works and when to let the LLM write SQL against your database instead.INTERMEDIATEMulti-Hop RAGYou'll understand how multi-hop RAG chains several retrieve-and-reason steps to answer questions a single lookup can't.ADVANCED