Use OpenSearch as Your Vector Search Database To Optimize GenAI Accuracy!
summary
Shows how OpenSearch becomes a production vector DB: enable the KNN plugin, store dense and sparse embeddings alongside inverted indexes, and use chunking strategies and hybrid queries to build RAG prompts. Walks through ingesting PDFs, creating embeddings with HuggingFace/OpenAI, indexing chunks, retrieving top vectors plus keyword hits, and appending those results to the LLM prompt. Claims enterprise tuning and architecture patterns yield 40%+ hallucination reduction in a real financial compliance chatbot.
AI-written summary · about
chapters (14)
- 0:00Optimizing GenAI with OpenSearch
- 2:29Goals and objectives
- 3:10Why vector search matters
- 7:23Vector embeddings in AI
- 9:10Vector search benefits for LLMs
- 10:15OpenSearch for vector workloads
- 12:48Vector search in OpenSearch
- 13:09What is RAG
- 15:18Implementing RAG with OpenSearch
- 17:47Optimizing OpenSearch for vectors
- 19:31Reducing hallucinations with use case
- 20:55What is NetApp InstaCluster
- 24:10Key takeaways and next steps
- 25:59Q&A
related talks
47mBoosting AI With OpenSearch: Vector Stores, Similarity Search, and Retrieval-Augme...
17mVectors Vs. Hallucinations: OpenSearch's GenAI Survival Kit
55mVector Search Made Simple: Getting Started with OpenSearch for AI Applications
22mVector Search Made Simple: Getting Started With OpenSearch for AI Applications
39mExperiments with OpenSearch and AI
26mOpenSearch: Conversations with Context
41mGenAI, vector/semantic/hybrid search, RRF, NLP, (...) and other buzzwords
48mEnter the Brave New World of GenAI with Vector Search