• Startsiden
  • Kurs
  • AI-500: Design and implement multi-agent AI solutions

AI-500: Design and implement multi-agent AI solutions

Ready to design and build production-grade multi-agent AI solutions? This advanced, four-day course teaches you how to architect scalable multi-agent systems using Microsoft Foundry and Azure. You'll gain hands-on experience designing the logical architecture behind multi-agent solutions, building and integrating tool ecosystems, implementing orchestration across agents, and weaving monitoring, security, and governance into your solutions.

Prerequisites

To get the most out of this course, you should have:

  • Completed AI-103 "Develop AI agents on Azure," or have equivalent hands-on experience building and deploying agents
  • Practical experience with Microsoft Foundry Agent Service and the Microsoft Agent Framework
  • Familiarity with common multi-agent orchestration patterns (sequential, concurrent, group chat, handoff)
  • A basic understanding of the A2A protocol and how to connect to remote agents
  • Experience consuming MCP tools and custom function tools
  • Working knowledge of RAG (Retrieval-Augmented Generation) patterns, ideally with Azure AI Search
  • Experience with system prompt design for shaping agent behavior and persona
  • Strong Python skills, including asynchronous programming and REST API consumption

Target audience

This course is aimed at experienced practitioners with deep expertise in designing, building, and optimizing scalable, production-ready multi-agent AI systems. Typical roles include AI engineers, developers, solution architects, and AI edge engineers who collaborate closely with ML engineers, platform engineers, data scientists, and business stakeholders to turn complex requirements into production-ready multi-agent solutions.

Se alle kurs innen dette fagområdet

 

Learning objectives:

  • Design scalable, cost-efficient multi-agent architectures, including advanced orchestration patterns such as hub-and-spoke and hierarchical supervisor models.
  • Build and integrate robust tool ecosystems, including custom MCP servers and advanced RAG pipelines.
  • Apply zero-trust security and governance principles to multi-agent solutions, including identity and secrets management.
  • Set up end-to-end monitoring and evaluation of multi-agent systems using OpenTelemetry and Azure Monitor.
  • Design human-in-the-loop workflows and manage troubleshooting and incident response in production.

Course modules:

Architecting production-grade multi-agent AI solutions in Azure

  • Introduction
  • Design stateful agentic loops, including run-status handling and context accumulation
  • Configure agent reflection cycles and session state persistence
  • Implement advanced orchestration patterns, including hub-and-spoke and hierarchical supervisor designs
  • Evaluate trade-offs across frameworks such as Semantic Kernel, LangGraph, AutoGen, and CrewAI
  • Apply task decomposition strategies, including prompt chaining and LLM-driven adaptive planning
  • Design agent handoff protocols for context-preserving transitions
  • Design enterprise-scale communication architectures between agents

Building production-grade multi-agent capabilities with Microsoft Foundry

  • Introduction
  • Design advanced prompting strategies, including multiturn reasoning and dynamic context injection
  • Build layered defenses against prompt injection in untrusted input environments
  • Design system prompt frameworks for agent persona and behavior control
  • Develop custom MCP servers with authentication and production-grade error handling
  • Implement dynamic tool selection, routing logic, and result validation
  • Govern tool ecosystems through dependency management and versioning
  • Implement advanced RAG pipelines, including hybrid search combining keyword and semantic ranking

Deploying and governing agentic AI solutions in Azure

  • Introduction
  • Secure multi-agent systems using zero-trust architecture principles
  • Apply per-agent managed identities with least-privilege access
  • Design authentication flows, including managed identity, on-behalf-of, and key-based patterns
  • Manage secrets lifecycle with Azure Key Vault, including rotation and customer-managed keys
  • Design network controls to prevent lateral movement
  • Configure multitenant data isolation and compliance controls
  • Scale responsible AI governance using Azure AI Content Safety

Monitoring, evaluating, and operating multi-agent AI solutions in Azure

  • Introduction
  • Implement distributed observability using OpenTelemetry and Azure Monitor
  • Design distributed tracing and structured logging for agent decision paths
  • Build anomaly detection for abnormal agent behavior
  • Design evaluation frameworks and success metrics for multi-agent outcomes
  • Build regression testing pipelines for behavioral drift detection
  • Optimize multi-agent performance and cost through model routing and caching strategies
  • Design human-in-the-loop systems, including escalation and approval workflows
  • Debug and respond to production incidents in multi-agent solutions

 

This course maps to the Microsoft Certified: Multi-Agent AI Solutions Expert (beta) certification. 

FAQ – AI-500: Design and implement multi-agent AI solutions

Hva koster kurset?
Prisen er 26 500 NOK for 4 dager. Kurset inkluderer digital kursdokumentasjon og hands-on labs. Lunsj og forfriskninger er inkludert ved klasseromskurs.

Hvor lenge varer kurset?
Kurset går over 4 dager fra kl. 09.00 til 16.00 alle dager.

Hvordan gjennomføres kurset?
Kurset gjennomføres med engelsk kursdokumentasjon og engelsktalende instruktør. Undervisningen kombinerer arkitektur, praktiske øvelser og hands-on arbeid med Microsoft Foundry, Azure og multi-agent AI-løsninger.

Hvem passer kurset for?
Kurset passer for erfarne tekniske fagpersoner som skal designe, bygge og optimalisere skalerbare multi-agent AI-løsninger for produksjonsmiljøer.

  • AI Engineers
  • Software Developers
  • Solution Architects
  • AI Edge Engineers
  • Platform Engineers
  • Data Scientists og tekniske AI-spesialister

Hvilke forkunnskaper kreves?
For å få fullt utbytte av kurset bør du ha gjennomført AI-103: Develop AI agents on Azure, eller ha tilsvarende praktisk erfaring med å bygge og deploye AI-agenter.

Hvilke tekniske forkunnskaper anbefales?
Det anbefales erfaring med Microsoft Foundry Agent Service, Microsoft Agent Framework, multi-agent orchestration patterns, A2A-protokollen, MCP tools, custom function tools, RAG-mønstre og Azure AI Search. Du bør også ha sterke Python-ferdigheter, inkludert asynkron programmering og bruk av REST API-er.

Hva lærer jeg i løpet av kurset?
Du lærer hvordan du designer og bygger produksjonsklare multi-agent AI-løsninger på Azure. Etter kurset vil du kunne:

  • Designe skalerbare og kostnadseffektive multi-agent arkitekturer
  • Implementere avanserte orchestration patterns som hub-and-spoke og hierarchical supervisor models
  • Bygge og integrere robuste tool ecosystems
  • Utvikle custom MCP servers og avanserte RAG pipelines
  • Sikre multi-agent systemer med zero-trust-prinsipper
  • Bruke managed identities, Key Vault og least privilege access
  • Sette opp observability med OpenTelemetry og Azure Monitor
  • Designe human-in-the-loop workflows og incident response for produksjon

Er kurset praktisk rettet?
Ja. Kurset er hands-on og fokuserer på praktisk design, bygging, drift og governance av multi-agent AI-løsninger i Azure.

Hvilke temaer dekkes i kurset?
Kurset dekker blant annet:

  • Produksjonsklar multi-agent arkitektur i Azure
  • Stateful agentic loops og session state persistence
  • Agent reflection cycles og context accumulation
  • Semantic Kernel, LangGraph, AutoGen og CrewAI
  • Prompt chaining og LLM-driven adaptive planning
  • Microsoft Foundry og Microsoft Agent Framework
  • Custom MCP servers, tool routing og result validation
  • Advanced RAG pipelines og hybrid search
  • Zero-trust security, identity og secrets management
  • Monitoring, evaluation, regression testing og drift

Får jeg sertifisering etter kurset?
Kurset mapper til sertifiseringen Microsoft Certified: Multi-Agent AI Solutions Expert (beta).

Er eksamen inkludert?
Kursinformasjonen oppgir digital kursdokumentasjon og hands-on labs som inkluderte elementer. Eksamen er ikke oppgitt som inkludert i kursprisen.

Hva gjør dette kurset avansert?
Kurset går utover grunnleggende agentutvikling og fokuserer på arkitektur, sikkerhet, governance, observability, skalering og produksjonsdrift av komplekse multi-agent AI-systemer.

Kan kurset leveres bedriftsinternt?
Ja. Kurset kan leveres både som åpent kurs og som bedriftsinternt kurs.


AI agents & Microsoft Foundry

Azure AI development

AI architecture, governance & secure development