THRSP1134 Theater Thu 19 Nov 13:40 PST
Trace, evaluate, and ship reliable AI agents
Peter Laudati, Jasmine Greenway
Session description
Microsoft's catalogue description
Catching a bad prompt, tool, or model change before it reaches production comes down to a repeatable loop: trace, evaluate, monitor, and test. Put this loop into practice with an agent deployed on Microsoft Foundry. Trace LLM and tool calls with Datadog Agent Observability to identify regressions and establish a repeatable framework for monitoring and governing AI applications at scale.