× × × × ×
× × × × ×
Designing and Implementing multi-agent AI Solutions
Course Description
AI-500: Designing and Implementing Multi-Agent AI Solutions
Architect and build production-grade multi-agent AI systems on Microsoft Foundry. This hands-on Azure course teaches you to design agent architectures, build tool ecosystems, orchestrate multiple agents, and add monitoring, security, and governance. It suits beginners and experienced architects alike, and it prepares you for the AI-500 exam.
Duration: 4 days (about 28-32 hrs) | Format: Instructor-led or self-paced | Level: Beginner to advanced | Prerequisites: Basic Python or C# and REST APIs
What You’ll Learn
- Design the logical architecture of a multi-agent solution
- Build and integrate tool ecosystems with function calling, OpenAPI, and MCP
- Implement multi-agent orchestration with the Microsoft Agent Framework
- Ground agents in enterprise data with RAG and Azure AI Search
- Add multimodal and document-processing capabilities to agents
- Apply monitoring, tracing, security, and governance to production agents
Who Is This Course For?
Solution architects, AI engineers, software developers, and technical leads. Beginners get an on-ramp module and guided labs. Experienced architects can focus on design, security, and governance content.
Course Curriculum
-
LLM basics
-
what agents are
-
Azure essentials
-
environment setup
-
Foundry projects
-
model deployments
-
Foundry Agent Service concepts (agents, threads, runs)
-
When to use one agent vs. many
-
agent roles and boundaries
-
logical architecture diagrams
-
reference patterns
-
cost and latency trade-offs
-
RAG fundamentals
-
Azure AI Search
-
vector and hybrid search
-
agentic retrieval
-
document extraction
-
Function calling
-
OpenAPI tools
-
built-in tools
-
Model Context Protocol (MCP) servers and clients
-
tool design best practices
-
Short-term (threads) and long-term memory
-
context management
-
state persistence
-
Microsoft Agent Framework
-
sequential, concurren
-
handoff
-
group chat
-
supervisor pattern
-
human-in-the-loop approvals
-
agent-to-agent interoperability
-
Vision
-
speech
-
document understanding inside agent workflows
-
Tracing across agents
-
Application Insights
-
token and cost tracking
-
agent evaluation
-
debugging failed runs
-
Entra ID and managed identity
-
least-privilege tool access
-
content safety
-
prompt-injection defenses
-
audit logging
-
policy and compliance
-
CI/CD
-
infrastructure as code
-
scaling
-
resiliency
-
API gateway patterns
-
continuous evaluation
-
Design and build a governed
-
monitored multi-agent assistant with grounded knowledge
-
MCP tools
-
human approval steps
-
an architecture document
Course Eligibility
Audience profile
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.
Audience prerequisites
For this course, you should have experience developing AI and machine learning, deploying agentic
systems in production environments, and orchestrating agent logic using Microsoft Foundry. You should
also be proficient in Python. Having experience in designing and implementing Azure or other cloud and
hybrid architecture designs is beneficial.
Reviews
