Skip to main content
Claude Partner Network - Certified

Gennoor is Certified in the Claude Partner Network.

Certifications & Projects · part of Learning Studio
‹ All Learning Studio courses

Microsoft, certification course, expert level

AI-500: Designing and Implementing Multi-Agent AI Solutions

Build a platform of cooperating AI agents on Microsoft Foundry in Python, and prepare for the AI-500 expert exam

You sign in once to start.

19
modules
441
video lessons
34
quizzes
19.9 h
of video
0:00 / 0:00
  • 139 lectures
  • 264 lab videos
  • 38 demos
  • 34 quizzes

Each bar is one lesson. Gaps separate the modules.

What you will be able to do

  • Design stateful agentic loops with the Responses API, checkpoints, forks and resume
  • Orchestrate teams of agents with hub-and-spoke, hierarchical and parallel patterns in the Microsoft Agent Framework
  • Decompose tasks and let agents collaborate and communicate with each other using A2A
  • Connect agents to tools and data with MCP servers, Azure AI Search retrieval and Azure Cosmos DB memory
  • Ship agent changes safely with GitHub Actions pipelines, versioning and rollback
  • Secure agents with zero-trust identity, and govern them with responsible AI controls and lifecycle policies
  • Trace, evaluate and budget a multi-agent solution, and add human approval where it matters
  • Prepare for Microsoft exam AI-500 across its skill areas

Curriculum

475 items (lessons and quizzes) across 19 modules. Pick any lesson to open it. You will sign in first if you have not already.

Course Introduction
5 lessons
Module 1: Design Stateful Agentic Loops
28 lessons
Module 2: Advanced Multi-Agent Orchestration
34 lessons
Module 3: Task Decomposition and Collaboration
26 lessons
Module 4: Agent-to-Agent Communication
27 lessons
Module 5: Advanced Prompting for Production Agents
30 lessons
Module 6: Tool Ecosystems with MCP
29 lessons
Module 7: Advanced RAG with AI Search
28 lessons
Module 8: Agent Memory with Cosmos DB
27 lessons
Module 9: CI-CD with GitHub Actions
27 lessons
Module 10: Zero-Trust Security
27 lessons
Module 11: Responsible AI Governance
26 lessons
Module 12: Agent Lifecycle Governance
27 lessons
Module 13: Observability Across Agents
27 lessons
Module 14: Evaluating Whole Claims
26 lessons
Module 15: Back Within Budget
26 lessons
Module 16: People in the Loop
28 lessons
Module 17: The 3-47 AM Incident
26 lessons
Course Wrap-up
1 lesson

About this course

Most teams run one AI agent. This course teaches you to design and build a whole system of agents, and to run it in production.

You join Larkspur Mutual, an insurer whose claims-intake agent is doing well, until the business asks for more than one agent can carry. Across seventeen modules you follow one company and one claim platform: you design stateful agent loops, coordinate teams of agents, connect them to tools and data, secure and govern them, measure them, control their cost, and handle the night something goes wrong.

What you build

  • Design

    stateful agentic loops with checkpoints and resume, orchestration patterns (hub and spoke, hierarchical, parallel), task decomposition and agent collaboration, and agent-to-agent communication with A2A.

  • Build

    production prompting for agents, tool ecosystems with MCP, advanced retrieval with Azure AI Search, agent memory in Azure Cosmos DB, and CI/CD with GitHub Actions.

  • Secure and govern

    zero-trust security for agents, responsible AI governance, and agent lifecycle governance from first version to retirement.

  • Operate

    observability across agents, evaluation of whole claims, staying within budget, human approval in the loop, and incident response at three forty-seven in the morning.

How every module works

  1. Story

    A short chapter sets up the problem

  2. Concept

    Built up on screen as a diagram

  3. Code

    Real code with the current SDK names

  4. Lab

    Every step run on camera

  5. Demo

    The same idea in the portal

A short Larkspur chapter sets up the problem. The concept builds up on screen as a diagram, with the design decision and its trade-offs. Then you see real Python code with the current SDK names. Hands-on lab videos follow, where every step is run on camera, and portal demos show the same ideas in the Azure portal and Microsoft Foundry.

What you get

19.9

hours of video

139

lectures

264

lab videos

38

portal demos

34

quizzes

  • 19.9 hours of video: 139 lectures, 264 lab videos and 38 demos
  • A hands-on lab in every module, built in parts that join into one working solution
  • lab code and sample data, lab guides, demo guides, knowledge checks and the slides of every lecture (PDF)
  • 34 quizzes that match the lectures and labs
  • English captions on every video
  • getting an Azure subscription, AI services on Azure, Microsoft Foundry and a model end to end

Exam AI-500 and beyond

The course is aligned to the skills measured for exam AI-500, which leads to the Microsoft Certified: Multi-Agent AI Solutions Expert certification. Where Microsoft is retiring a feature, you learn its successor, so what you build matches what is current today.

By the end you can design, build, secure, evaluate and operate multi-agent AI solutions on Azure, and you are ready to prepare for the exam with confidence.