Agentic coding with GitHub Copilot
Are you ready to go beyond simple chat-based coding and unlock the power of structured, agentic workflows? In this hands-on workshop, you'll learn how to move beyond chat-based coding and design structured, agentic coding workflows. Using GitHub Copilot, you will learn to write effective prompts, leverage LLMs for multi-file coding tasks, and design workflows that plan, act, and iterate safely. You’ll also explore how to evaluate agent outputs, implement guardrails, and address important risks and governance issues. By the end of the training, you'll be able to use LLMs for structured coding tasks, design robust agentic workflows, and enhance productivity in your own projects.
Learning outcomes
- Explain key concepts and terminology related to agentic coding (agents, planning, loops, tools, guardrails)
- Describe the agent loop (goal → plan → act → observe → revise) and its implications for software development
- Summarize common failure modes and risks of coding agents
- Write effective prompts and task descriptions that guide agent behavior
- Critically review agent-generated code for correctness, maintainability, and security
- Design an end-to-end agentic workflow for a realistic coding scenario
Approach
The course features a mix of explanations, live demonstrations, and hands-on exercises, giving you the opportunity to apply LLMs to real-world coding challenges.
To follow along effectively with this course and do the exercises, you need to have a license for GitHub Copilot.
Required skills
- Basic programming experience (Python or other)
- Familiarity with chat-based LLM tools
- Familiarity with Git is advisable
- GitHub account (and license)
Software demonstrated
- VS Code
- GitHub (Copilot)
- Python and R
Extra information
This course is part of a learning path Machine learning.