WHOOP builds fitness wearables and a membership platform that helps people train smarter. Their devices capture biometric data 24/7 and translate it into guidance on strain, recovery, and sleep for members from everyday runners to world-class athletes like Christiano Ronaldo.
When WHOOP launched its AI coach, expectations changed. Members need coaching that is fast, consistent, and trustworthy. WHOOP’s AI and engineering teams also need to iterate quickly without quality drift, latency spikes, or cost surprises in production.
In this session, the WHOOP team shares how they evaluate and improve the coach using Datadog LLM Observability: tracing each request end to end, turning real traffic into evaluation datasets, and running structured experiments to compare prompt and model changes before rollout.
If you are building agents and LLM applications, you won’t want to miss this. You’ll leave with a playbook to define success metrics, catch regressions early, and ship safer updates that users keep trusting.
Speakers
Speakers
Speakers