Agentic AI for Scientists: A 6-Week Hands-On Workshop Series

Description

Agentic AI for Scientists is a 6-week hands-on workshop series designed to introduce researchers and engineers to the rapidly evolving landscape of agentic AI. Beginning with the foundations of modern AI, the series traces the progression from classical machine learning and deep learning to large language models and, ultimately, agentic AI systems. Moving from foundational concepts to advanced applications, the workshop explores agent design patterns, multi-agent architectures, post-training, deployment, evaluation, and the emerging frontier of autonomous AI agents for scientific research.

Designed for participants with diverse backgrounds and levels of experience, the series provides both a strong conceptual understanding of AI agents and practical insight into how they can be applied to scientific workflows. Participants explore how agentic AI can support research automation, data analysis, and knowledge discovery, while gaining hands-on experience through weekly projects and applied exercises. Every delivered week is published openly as runnable Colab notebooks, slides, and notes, built on top of Organon, an agent-first operating system for scientists.

The six weeks

• Week 1, GenAI Foundations for Scientists. A brief history of the AI revolution, from machine learning and deep learning to Transformers, large language models, and AI agents, plus a practical introduction to coding agents.
• Week 2, AI Agents Fundamentals I: Patterns. Core agentic patterns: ReAct, CoT/ToT-style reasoning, and reflection agents, with tool use, function calling, and retrieval-augmented grounding (RAG).
• Week 3, AI Agents Fundamentals II: Multi-Agent Systems. Designing multi-agent systems for deep research and automated knowledge discovery.
• Week 4, Post-Training and Deployment of AI Agents. How to shape model behavior and prepare agents for real-world use.
• Week 5, Evaluation and Benchmarking of AI Agents. Measuring whether an agent is reliable and good enough to trust.
• Week 6, Towards Full Autonomy. The emerging frontier of autonomous AI agents for scientific work.

What participants learn

• How modern AI evolved from machine learning to large language models and AI agents.
• The core design patterns behind agentic systems, including ReAct, reasoning-based approaches, and reflection-style agents.
• How agents use tools, external APIs, and retrieval systems such as RAG to work with real information.
• How multi-agent systems can be designed for research automation, deep research, and knowledge discovery.
• What post-training and deployment involve, and how to evaluate and benchmark agents for real-world scientific use cases.

The series is led by Dr.-Ing. Kerem Delikoyun, a research fellow at TUMCREATE working at the intersection of biomedical imaging, translational medical technology, and agentic AI. Each session also features special guest speakers with domain expertise relevant to the week's topic.