From embeddings and chunking through hybrid retrieval, reranking, GraphRAG, multimodal RAG, agentic RAG, evaluation, and production deployment. 16 lessons covering the full Perplexity/Glean-style production stack.
Grounding LLMs in facts. Why RAG beats raw LLMs for knowledge-heavy tasks. The retrieve-augment-generate flow.
Turn text into numbers. Dense vs sparse embeddings, cosine similarity, fine-tuning embeddings for domain.
Pinecone, Weaviate, Chroma, Qdrant, pgvector, Milvus — choosing and operating production vector storage.
Greg Kamradt's 5 levels: char → recursive → document-specific → semantic → agentic. Late-chunking and hierarchical.
End-to-end RAG pipeline walkthrough — document loading, chunking, embedding, retrieval, prompt assembly, and generation. How failures cascade.
Dense, sparse (BM25), hybrid via RRF, MMR for diversity. The retrieval algorithm taxonomy.
HNSW, IVF-PQ, ScaNN, DiskANN — billion-vector serving with the right index, sharding, and recall targets.
Cross-encoder rerankers (Cohere Rerank, BGE), ColBERT late interaction. Two-stage retrieve-then-rerank.
HyDE, Step-Back, RAG-Fusion, multi-query, sub-question decomposition. Query-side techniques.
Self-RAG, CRAG, Adaptive RAG. Retrieval critics and multi-hop loops.
Microsoft GraphRAG, Neo4j integration, entity/community summaries — when graphs beat vectors.
ColPali late-interaction VLM, ViDoRe benchmark, image+text+table retrieval for visually rich documents.
RAGAS metrics (faithfulness, answer relevancy, context precision/recall), CRAG benchmark, golden sets, citation-grounded eval.
Gemini 2M / Claude 1M decision rubric. U-shaped attention, cost ratios, hybrid patterns.
Prompt injection via retrieved content, semantic caching, serverless vector infra, observability, cost optimization.
Capstone — Perplexity/Glean-style end-to-end app. FastAPI + hybrid retrieval + reranker + streaming + citations + evaluation + monitoring.
9 interactive labs — hands-on exercises for this track
A customer asks about your product — watch the AI search docs, find answers, and cite sources
Drag a query star across an embedding map and watch nearest neighbors update instantly
See how Fixed-Size, Sentence, Paragraph, and Semantic chunking split documents for RAG
Drag sentences, type queries, and watch nearest-neighbor retrieval light up live. See the geometry behind RAG in 2D.
Watch a cross-encoder reranker pull a buried answer from rank 7 to rank 1 after vector search.
See how BM25, vector search, and reciprocal rank fusion rank the same documents — different queries favor different methods.
Drag K and watch Precision@K, Recall@K, and MRR update live on three pre-labeled retrievers.
See how short user queries get rewritten by HyDE and multi-query expansion — and how retrieval quality changes.
20k-token doc vs 8k context. 5 strategies (truncate, sliding, summarize, map-reduce, recursive) compared on cost vs answer recovery.
400 questions across 16 modules — check how well you understood this track.