Personalisation and Specialisation of Search With OpenSearch Agentic Search
summary
Demo of a secondhand luxury store shows the query "dress watch" rewritten by an OpenSearch flow agent with a query planner into OpenSearch DSL so results list watches not dresses. Uses Qwen 3 4‑bit (~<6GB) hosted via Olama/vLLM or cloud (Bedrock/SageMaker); baseline retrieval is BM25 with optional hybrid vectors and reranking. Claims fine‑tuning via Axolotl/LLaMA‑Factory improves rewrite quality; addresses eCommerce latency needs and prompt caching to meet sub‑100ms retail SLA, and uses persona contexts (Kami €1,500 budget, Julian collector, Sophia stylist).
AI-written summary · about
chapters (15)
- 0:00Agentic search in OpenSearch
- 1:07Keyword search limitations demo
- 2:11Intent-based results with agentic search
- 3:35Personalizing search for personas
- 6:47Search relevancy challenges and fixes
- 9:13Actions to improve search quality
- 10:41Data optimization for search
- 11:22OpenSearch overview
- 13:26Agentic search mechanics
- 16:19Fine-tuning for specialized queries
- 19:30Training data for fine-tuning
- 20:13Agentic search workflow setup
- 21:17Flow vs conversational agent
- 24:50Q&A on agentic search performance
- 28:46Q&A on fine-tuning and deployment
related talks
21mPersonalize Search Results with OpenSearch Agentic Memory
22mAgentic Hybrid Retrieval: Orchestrating LLM Agents, Hybrid Search & Explainability in OpenSearch
40mAgentic Relevance Tuning With OpenSearch: Faster, Easier, and Higher Quality Search at Scale
36mBuilding Next-gen Search With OpenSearch
37mAgentic AI Meets Search: Real-World Use Cases With...
40mBuilding Intelligent Search Systems With Agentic RAG and OpenSearch
26mRevolutionizing Search: Unveiling Conversational Search
38mAI-Ready Data Foundation Using Amazon OpenSearch