This page lists exercises associated with Microsoft skilling content on Microsoft Learn
Design stateful agentic loops with Microsoft Foundry Agent Service
Level: 400 | Duration: 45
Implement a bounded Agents v2 reflection loop, persistent conversation, and resumable branch for investment research.
Implement advanced multi-agent orchestration patterns in Microsoft Foundry
Level: 400 | Duration: 45
Build and validate a v2 fan-out/fan-in research workflow with quorum and partial-failure handling.
Apply task decomposition and agent collaboration strategies in Microsoft Foundry
Level: 400 | Duration: 45
Implement a live meta-agent planner, validated task DAG, and context-preserving specialist handoffs.
Design enterprise-scale agent communication with A2A in Azure
Level: 400 | Duration: 45
Implement tenant-isolated A2A discovery, JSON-RPC messaging, Cosmos DB shared state, and conflict audit trails.
Design advanced prompting strategies for production AI agents
Level: 400 | Duration: 45
Implement versioned clinical prompts, four-surface guardrails, live A/B evidence, and fine-tuning data preparation.
Build an enterprise MCP tool ecosystem
Level: 400 | Duration: 45
Implement, discover, invoke, and govern clinical tools through a real MCP server and client.
Implement advanced RAG with Azure AI Search
Level: 400 | Duration: 45
Build and validate a routed hybrid, vector, and semantic retrieval pipeline against live Azure AI Search.
Build persistent shared memory for Foundry agents
Level: 400 | Duration: 45
Build a Microsoft Foundry Hosted Agent that recalls patient-scoped vector memory from Azure Cosmos DB with retention, context budgeting, audit, and consistency controls.
Implement CI/CD for Foundry hosted agents
Level: 400 | Duration: 45
Release immutable Microsoft Foundry hosted-agent versions with GitHub Actions, then use a thin Container Apps dashboard to demonstrate progressive delivery and compatible-set rollback.
Apply zero trust to a Microsoft Foundry multi-agent workflow
Level: 400 | Duration: 45
Secure a Foundry code-review agent graph with identity-scoped access, tenant propagation, lateral-movement prevention, data minimization, and compliance evidence.
Govern a Foundry multi-agent code review
Level: 400 | Duration: 45
Apply content safety, fairness, transparency, privacy, and accountability controls to a Microsoft Foundry multi-agent workflow.
Govern a multi-agent topology lifecycle
Level: 400 | Duration: 45
Use Microsoft Foundry and Cosmos DB interfaces to govern agent versions, quotas, rate limits, metering, chargeback, and retirement.
Trace multi-agent workflows with OpenTelemetry
Level: 400 | Duration: 45
Create correlated OpenTelemetry spans, W3C context propagation, structured logs, anomaly signals, and Azure Monitor evidence for Adventure Works agents.
Build a multi-agent evaluation quality gate
Level: 400 | Duration: 45
Evaluate Adventure Works multi-agent journeys with Microsoft Foundry evaluators, synthetic data, calibration evidence, and regression thresholds.
Optimize multi-agent performance and cost with measured evidence
Level: 400 | Duration: 45
Measure live model usage, then implement model routing, caching, token budgets, and quality-floor escalation for Adventure Works.
Implement durable human approval workflows
Level: 400 | Duration: 45
Build a resumable Adventure Works approval workflow with calibrated escalation, Service Bus events, Cosmos DB state, and immutable audit evidence.
Debug a multi-agent production incident with trace evidence
Level: 400 | Duration: 45
Use Application Insights, KQL, trace snapshots, safe replay, structured hypotheses, and evidence-linked postmortems to diagnose an Adventure Works incident.