Optimizing Context Tuning for RAG Systems
summary
Shows concrete outcomes: a data pipeline with automated updates, monitoring and chunking experiments that compare fixed-size chunking versus recursive, semantic chunking and agentic chunking; compares BM25 sparse, dense (embedding) and hybrid retrievers using Pinecone and Chroma/FAISS-style indexes; adds cross-encoder reranking, time-decay and source-authority weights; generates evaluation QA pairs with an LLM and measures precision@K to iterate. Reports tradeoffs in cost, latency and context-window loss and when each strategy improves end-to-end QA accuracy.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:07What is RAG
- 3:38RAG data formats
- 5:25Model management in RAG
- 6:27Text chunking strategies
- 10:21Retrieval strategies
- 13:10Named entity recognition
- 15:19Context augmentation
- 18:53Evaluating RAG performance
- 20:40Closing remarks
- 21:22Q&A: Large document limits
- 23:28Q&A: RAG future
- 24:33Q&A: Preventing chunk splits
related talks
27mOptimizing RAG Systems in Practice: Challenges and Proven Solutions
15mEvaluating RAG Model Performance Metrics, Bias, and Interpretability
25mAdvanced RAG Systems: Scaling Retrieval-Augmented Generation in Production
19mRAG in 2025: State of the Art and the Road Forward
28mImproving responses in generative AI applications
30mI Want to Build a RAG System, Now What?
59mThe Evolution of RAG
49mContext is King