Design and implement multi-agent AI solutions

By the end of this module, you'll be able to:

• Design agentic loop event handling for production run lifecycles
• Implement reflection and planning cycles for multi-step reasoning
• Architect session state persistence for long-lived workflows
• Build fork-based session patterns for workflow branching
• Describe the Agents v2 runtime model including agents, conversations, responses, and items
• Migrate stateful agentic loop code from Agents v1 to Agents v2 using the azure-ai-projects 2.x SDK
• Design distributed tracing architecture that links all agent operations into a single correlated trace.
• Implement structured logging that captures agent decision points in a queryable format.
• Configure telemetry aggregation and operational dashboards that reveal system-wide patterns.
• Build anomaly detection for agent behavioral drift.

Candidates for this course, are expert-level practitioners who have subject matter expertise in designing, building, and optimizing scalable, production-ready, multi-agent AI?systems, solutions and workflows. Out of your job role, you work closely with developers, machine learning engineers, platform engineers, data scientists, and business stakeholders to translate complex requirements into production-ready, multi-agent solutions.
 

• Completion of AI-103 'Develop AI agents on Azure' or equivalent hands-on experience building and deploying agents
• Working experience with Microsoft Foundry Agent Service and Microsoft Agent Framework
• Familiarity with foundational multi-agent orchestration patterns (sequential, concurrent, group chat, handoff)
• Basic knowledge of the A2A protocol and how to connect to a remote agent
• Python programming proficiency, including async patterns and REST API consumption
• Familiarity with classic RAG retrieval patterns using Azure AI Search (hybrid search, semantic ranking); experience with Foundry IQ agentic retrieval is beneficial but not required
• Experience with system prompt design for agent persona and behavior control
• Python proficiency with the Azure AI SDK and Azure OpenAI SDK
• Working experience with GitHub Actions for CI/CD pipeline design
• Basic knowledge of Azure security services (Azure Key Vault, Microsoft Entra ID)
• Familiarity with responsible AI principles, Azure AI Content Safety guardrails, and basic Microsoft Entra ID concepts (managed identities, RBAC)

• Architect production-grade multi-agent AI solutions in Azure
  • Design stateful agentic loops with Microsoft Foundry agent service
  • Implement advanced multi-agent orchestration patterns in Microsoft Foundry
  • Apply task decomposition and agent collaboration strategies in Microsoft Foundry
  • Design enterprise-scale agent communication with A2A in Azure
• Build production-grade multi-agent capabilities with Microsoft Foundry
  • Design advanced prompting strategies for production AI agents
  • Build enterprise-grade tool ecosystems with MCP and Microsoft Foundry
  • Implement advanced RAG pipelines with Azure AI Search and Microsoft Foundry
  • Design multi-agent memory architectures with Azure Cosmos DB
• Deploy and govern enterprise agentic AI solutions on Azure
  • Implement CI/CD pipelines for multi-agent systems with GitHub Actions
  • Secure multi-agent systems with Azure zero-trust architecture
  • Scale responsible AI governance with Azure AI Content Safety and Microsoft Foundry
  • Govern the enterprise agent lifecycle in Microsoft Foundry
• Monitor, evaluate, and operate multi-agent AI solutions in Azure
  • Implement distributed observability for multi-agent solutions with OpenTelemetry
  • Design evaluation frameworks for multi-agent solutions with Microsoft Foundry
  • Optimize multi-agent performance and cost in Microsoft Foundry
  • Design human-in-the-loop approval workflows with Power Automate and Microsoft Teams
  • Debug and respond to production multi-agent incidents in Azure