From RAG to Riches: Secrets of Data-Augmented LLMs
summary
Explains Retrieval-Augmented Generation (RAG) as combining large language models with traditional search, and examines practical promises and pitfalls like hallucination. Covers concrete techniques—query reformulation, scoring, reranking, context fusion—and how to orchestrate these components into robust, usable RAG pipelines for developers and architects without requiring huge compute.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:15Speaker context
- 3:03Generative AI and LLMs
- 7:20LLM problems and RAG
- 14:48Semantic embeddings and RAG
- 20:28Naive RAG pipeline
- 24:22Naive RAG limits
- 27:09Document corpus challenges
- 31:43Chunking for coherence
- 36:35User question issues
- 40:49Similarity search relevance
- 44:27Advanced RAG techniques
- 46:13RAG observability
- 50:11LLM judge evaluation
- 51:56Conclusion
related talks
44mWhat Is a RAG System and How to Build One
51mDeveloping a Custom RAG System
01h 33mRAG How To Talk To Your Data
44mRAG for a Medical Company: Technical and Product Challenges AI & Data
38mRetrieval-Augmented Generation (RAG) in NLP
28mNavigating Challenges and Enhancing Performance of LLM based Applications
57mLLMs gone wild
42mTo RAG or not to RAG