RAG: from dumb implementation to serious results
summary
Shows concrete fixes and code using LangChain4j and Gemini/Vertex AI: embedding sentences but returning paragraph context with a sliding-window sentence embeddings trick, boosting similarity from about 0.91 to 0.94 and to 0.95 with hypothetical targets; generating hypothetical-question embeddings during ingestion to move the true-answer chunk from 4th to 1st position; practical metadata filtering, reranking, query compression, query routing, function-calling agents, and evaluation advice plus reminders to update stale vectors.
AI-written summary · about
chapters (10)
- 0:00RAG introduction
- 8:30Document chunking
- 17:53Hypothetical questions
- 21:48Contextual retrieval
- 28:52Query compression
- 33:13Reranking techniques
- 38:12Agentic RAG
- 44:11RAG evaluation & security
- 45:36Multi-hop questions
- 47:25Vector DB vs context window
related talks
56mRAG: from dumb implementation to serious results
53mBuild RAG from Scratch
30mRAG: Tricks from the trenches.
48mCrafting your own RAG system: Leveraging 30+ LLMs for enhanced performance
35mThe Hidden Jewels to Solve Confabulation Challenges in Generative AI’s RAG Promise
01h 33mRAG How To Talk To Your Data
44mBuild RAG from Scratch
48mGoing Beyond Prompt Engineering