Build and Scale AI Agents with Microsoft Foundry

This lab is the most advanced of three Microsoft Foundry agent-building learning paths, aimed at developers, engineers and architects who want full control over models, orchestration and operations rather than a no-code or low-code approach. Across five challenges, participants create purpose-built agents with system prompts, tools and domain-specific data, instrument them with OpenTelemetry-based tracing through Application Insights, run LLM-as-judge evaluations to measure output quality, and wire agents into automated multi-agent workflows using the Python SDK and the Foundry portal.

Three parallel scenarios cover the same Foundry concepts in different settings: detecting machine anomalies at a fictional manufacturer, triaging insurance claims, and classifying call-center intents, and the hackathon mixes writing and running Python with using the Foundry portal to deploy models and review traces and evaluations.