RAGs to Riches: How Our Content Affects Retrieval Augmented Generation
summary
Claims that RAG cuts hallucinations and speeds doc drafting — a vector-based RAG plus templates and on‑the‑fly ingestion for uploads powers internal verba GPT, integrated with a CICD pipeline and privacy‑first inference. Implements custom chunking, embeddings, a vector DB, and delta updates to keep content current. Results: first‑draft turnaround drops from two weeks to two days and edit drafts from two days to five minutes; next steps add graph RAG and agentic tooling.
AI-written summary · about
chapters (16)
- 0:00RAG and technical communicators
- 1:07LLM limitations and hallucinations
- 3:44What is retrieval augmented generation
- 5:08Vector-based RAG workflow
- 7:17Vector-based RAG limitations
- 8:03Document processing for RAG
- 13:18Graph-based RAG and knowledge graphs
- 15:28Graph-based RAG challenges
- 17:16Optimizing docs for RAG
- 19:22Skyflow's RAG implementation
- 24:02Future RAG trends
- 26:41Ethics in AI content
- 28:08Technical communicators' evolving role
- 30:17Q&A: Keeping AI content current
- 32:44Q&A: Starting with RAG
- 35:57Q&A: Teams for RAG adoption
related talks
42mTo RAG or not to RAG
51mDeveloping a Custom RAG System
50mRetrieval Augmented Generation Beyond Toy Demos
20mRetrieval Augmented Generation for AI Apps
25mAdvanced RAG Systems: Scaling Retrieval-Augmented Generation in Production
38mRetrieval-Augmented Generation (RAG) in NLP
22mThe Power of RAG at Scale: A Year of Explosive Growth and Observab...
19mImproving RAG Systems - Key Challenges and Solutions