Developing in Agent-Based AI Systems (GH-600)
Developing Reliable Multi-Agent Systems on GitHub – The Official Microsoft Training Course
.svg)
Machine Learning Engineer
You'll learn how to evaluate and optimize agent performance, implement governance measures and security safeguards, and coordinate multi-agent systems to ensure secure, reliable, and efficient results. Through hands-on learning, you’ll acquire the skills needed to operate, monitor, and control AI agents in production environments using GitHub as a control layer.
1. Fundamentals of Agent-Based AI on GitHub
Learn how AI coding agents are transforming software development by planning, acting, and making improvements within GitHub workflows.
2. Design of the Agent Architecture and Integration into the SDLC
Learn how agent-based systems use GitHub Workflows to develop software securely.
3. Tools, MCP, and Runtime Environments for Agents
Learn how agents use tools, MCP, and GitHub workflows to securely perform tasks—with clear boundaries, security controls, and scalable automation.
4. Multi-Agent Systems and Orchestration
Learn how to design reliable multi-agent systems in GitHub using observable workflows, coordinated artifacts, and secure recovery mechanisms.
5. Storage, Status, and Analysis
Learn how to manage agent memory and state, maintain progress across different environments, and evaluate agent behavior based on clear success signals.
6. Governance, Safeguards, and Operations
In this module, you'll learn how to design secure and compliant agent governance using GitHub's built-in controls, "human-in-the-loop" approvals, and the principle of least privilege. You'll also learn about operational safeguards to improve reliability, accountability, and recovery.
Requirements:
- A GitHub account
- A Basic Understanding of the Fundamentals of AI
- Basic Understanding of Repositories, Branches, and Pull Requests
- General knowledge of CI and CD concepts
- Integrate AI agents into the software development life cycle (SDLC) by defining agent tasks, inputs, outputs, and execution constraints
- Design and configure agent architectures that separate planning, reasoning, and execution to improve reliability and control
- Implement tool usage and interactions with the environment by configuring agent tools, permissions, and MCP servers within development environments
- Design reliable multi-agent systems on GitHub using observable workflows, coordinated artifacts, and secure recovery mechanisms
- Learn how to manage agent memory and state, maintain progress across environments, and evaluate agent behavior based on clear success signals
- Develop secure and compliant agent management using GitHub-native controls, "human-in-the-loop" approvals, and the principle of least privilege
This course consists of antraining is led by an instructor who provides live guidance to participants. Theory and practice are taught through live demonstrations and hands-on exercises. The course uses the video conferencing software Zoom.
You should have expertise in the operation, integration, monitoring, and control of AI agents within production-ready SDLC workflows and development environments, ensuring reliability, security, and speed, with GitHub serving as the system of record and control plane. You’ll work closely with architects, platformengineers, DevOpsengineers, application developers, product managers, and securityengineers to develop, deploy, operate, and manage agents that run within the GitHub platform. You should have experience with the software development life cycle (SDLC), GitHub workflows and controls, as well as practices related to code quality, security, and review. You should also have experience programming agents, including GitHub Copilot, MCP servers, and agent customization—such as custom instructions, custom agents, tools, and Copilot setup.
Your responsibilities will include:
- Running Agent Workflows Within the SDLC
- Monitoring Autonomous Behavior Using GitHub Checks
- Analysis and Optimization of Agent Results Based on Scans and Artifacts
- Configuring Custom Agents
- Reliable Coordination of the Execution of Multiple Agents
.svg)