19.05.2026 –, Raum C
Ever wished your AI could do more than just answer a question?
What if you could teach it how to think, decide, and verify?
This workshop introduces one of the newest and most powerful AI architectures: graph-based reasoning systems, where AI agents follow structured decision graphs to retrieve information, reason step by step, and produce reliable, evidence-based answers.
We begin from first principles, explaining how large language models work and why they cannot be trusted on their own. You will then build a complete RAG pipeline that connects an open-source LLM to real documents, enabling it to retrieve evidence and generate fact-grounded answers instead of confident guesses.
From there, we move beyond linear pipelines.
You will transform your RAG system into a graph-based AI agent: an autonomous assistant whose reasoning is explicit, structured, and controllable. Using a reasoning graph, the AI learns when to retrieve information, how to combine multiple sources, when to verify results, and when to stop. Instead of one-shot responses, your system follows a clear decision flow that mirrors human problem-solving.
By the end of the workshop, you will have built a complete, open-source AI assistant that can read documents, retrieve knowledge, reason through a graph, and answer with evidence.
Not a simple chatbot.
A truly intelligent AI agent that follows a path of thought.
What you will learn:
Master the Fundamentals : Learn how Retrieval-Augmented Generation (RAG) and AI agents work together to combine retrieval, reasoning, and generation into accurate, context-aware responses.
Build Your Own AI Agent System: Follow a step-by-step process to build the full pipeline from document ingestion, chunking, embeddings, retrieval, to LLM integration, and then upgrade it into a graph-based agent that knows when to retrieve, what to search, and how to produce grounded answers.
Gain Hands-On Experience: Work through practical exercises and real-world examples that show how a graph-based agent can explore different type of documents, gather evidence, verify key points, and answer in a reliable way.
Ensure Data Security: Learn how to deploy open-source AI systems with a strong focus on data privacy, secure processing, and responsible use of sensitive information.
Bonus: Evaluation & Reliability: Discover how to test and measure your AI system to ensure it is using the right sources, avoiding hallucinations, and producing reliable, trustworthy answers.
Who Should Join
This workshop is designed for anyone familiar with Python who wants to go deeper into how the most robust and modern AI systems are built. Perfect for:
💻 Developers
📊Data Scientists
🛠️Engineers
🤖Curious AI Enthusiasts
Some experience with Python programming is required. You don’t need any background in machine learning or artificial intelligence. We’ll build everything together, step by step.
What to Bring
Just bring your laptop and a stable internet connection (Wi-Fi). We’ll take care of the rest.
Why This Matters
Most LLMs are impressive… but also a little dangerous. They often sound confident even when they are wrong.
In this workshop, you will build an AI system where RAG provides knowledge, AI agents add judgment and graph-based reasoning allows your AI to search, verify, and think (in the right way) before it answers.
Ornella Vaccarelli is a Senior Research Scientist at iCoSys and the Lead Scientist at SCAI (Swiss Center for Augmented Intelligence), where she pioneers innovative AI solutions across diverse domains. With expertise ranging from computer vision and computational physics to the latest developments in large language models (LLMs), she bridges cutting-edge research and practical application.
Her collaborative projects span from fundamental research at EPFL to developing an LLM-RAG system for the Swiss parliament. Ornella’s work not only advances scientific understanding but also transforms how industry and government leverage AI for informed decision-making.
A prolific researcher, her findings have been published in high-impact journals, and she is a regular speaker at international conferences. Ornella earned her PhD in Computational Physics from Sorbonne University in Paris and holds a Master’s in Theoretical Physics from the University of Bari. Her career exemplifies a commitment to pushing the boundaries of AI while ensuring its responsible and effective integration into real-world applications.