Your Agents Are Searching. Is Your Search Learning?

Agenda:

• 7-7:30 pm: Networking and food/drinks
• 7:30-8:15 pm: Announcements and presentation
• 8:15 pm-8:30 pm: Q&A and close-out

Your Agents Are Searching. Is Your Search Learning?
Building an Autonomous Search Relevance Engine on OpenSearch

Your agent searches, and searches again, before it finally finds what it needs. Then it forgets. The next time, it starts over and burns the same tokens on the same dead ends. Your search saw every one of those attempts and learned nothing.
You need search that learns. In this talk we'll build the loop behind an autonomous search relevance engine, live on OpenSearch, with a real production index:

1. Record: capture every search and click with User Behavior Insights (UBI).
2. Find: turn that data into a scorecard of zero-result queries, weak results, and retries.
3. Fix: choose a change that targets the failure, not a guess.
4. Test: prove it offline against judged queries before anyone sees it.
5. Ship: roll it out to production, and watch the failures drop.

You'll walk away with a roadmap for self-improving search, and with a basic loop you can reuse for any autonomous development system: observe, diagnose, change, verify, ship.