Hybrid search based on OpenSearch and Qdrant
summary
Describes implementing hybrid search combining OpenSearch full-text search with Qdrant vector search at T-Bank, outlining the engineering problems encountered (integration, relevance blending, infrastructure) and the practical results achieved from deploying the combined system.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:17What is hybrid search
- 8:39Problem and first attempt
- 12:56Choosing vector database
- 16:51Filtering and postprocessing issues
- 20:04Optimizing with Rust
- 23:01Model training and caching
- 26:12Product domain application
- 30:45Scaling architecture
- 32:31Summary
- 32:53Q&A
related talks
46mLive coding a quick & easy semantic search with Qdrant from scratch
43mOpenSearch and the Path to a Trillion-scale Vector Database
41mText to Context: How We Introduced Hybrid Search
40mAgentic Relevance Tuning With OpenSearch: Faster, Easier, and Higher Quality Search at Scale
30mSearch on Steroids — A Story of Semantics
26mOpenSearch and the Path to a Trillion-s...
48mUnlocking Next-gen Semantic Search With OpenSearch 3.x
18mAdvancing Search at Uber with OpenSearch