Online workshop: Build a custom AI observability layer in Braintrust
The signals that reveal how well your AI is serving users are different for every product. Braintrust lets you define them as custom facets and track every production trace against them.
In this session, Nick Slavin shows how to design custom facets for your specific use case, use SQL-queryable trace metadata for trend analysis and reporting, and build a quality practice that catches failures and surfaces what's working.
What you'll learn
How to define the dimensions that matter to your business as custom Topics facets and cluster every trace against them automatically
What SQL-queryable trace metadata unlocks for analysis and reporting
How to track trends over time and know when a fix actually held
How teams scale production observability across multiple agents and product surfaces