Agentic AI goes beyond LLM-driven workflows: instead of defining how to do a task, an agent can plan and decide which tools to use in order to complete multi-step tasks. In this hands-on masterclass, we unpack what makes agents different, when they are the right solution (and when they are not), and how to build agentics systems in Python using modern patterns such as ReAct and Plan-and-Execute.
By the end of this masterclass, you will be able to:
- Explain agents vs. workflows and choose the right approach for a use case
- Describe the core building blocks (LLM, tools, memory) and the agentic loop
- Implement ReAct and Plan-and-Execute agents and extend them with structured tool use
- Engineer tools: design signatures and write model-friendly docstrings
- Package tool access behind the Model Context Protocol (MCP) and connect it to the agent
- Add memory to improve multi-step performance
- Outline orchestration and governance needs for developing AI Agents in production
Who is this masterclass for?
Python developers and data/ML practitioners who have basic experience using LLMs and want to move from prompts and workflows to building agentic systems.
Requirements
- Laptop with internet access
- GitHub account (to run the workshop in GitHub Codespaces)
- Comfortable with Python fundamentals
- 6 hours total, combining short conceptual segments with guided coding
- Hands-on implementation of a ReAct agent, then incremental extensions
- Practical discussions on suitability, safety boundaries, etc.
Trainers
The masterclass is created by the appliedAI Institute for Europe and will be given by Maternus Herold and Jan Willem Kleinrouweler.