In this introduction training we will explore the basic concepts and building blocks behind knowledge graphs: nodes, edges, triples, IRIs. We will focus on RDF-based knowledge graphs and the surrounding W3C stack (RDF, RDFS, OWL, SHACL, SPARQL), while also contrasting them with labelled property graphs such as Neo4j so that attendees understand the trade-offs between the two paradigms.

‍

From there we will move into ontology modelling: how to design a schema that captures a domain, reuse existing vocabularies and ontologies, and validate data. We will review the architectures and stacks commonly deployed in KG based systems such as triplestores, federated SPARQL endpoints, ETL and mapping layers, and graph RAG systems.

‍

We will then dive into use cases from several domains and give an overview of how knowledge graphs are leveraged, together with the tooling that supports this work. Finally, we will look at how knowledge graphs and LLMs complement each other, and guide attendees on using LLMs for knowledge base design and population such as ontology drafting, entity and relation extraction, triple generation, and validation-in-the-loop. Lastly, we will explore how graphs can ground LLM outputs in return.

This workshop includes hands-on sessions to:

-Create and develop and model for entities and properties, in a toy example,

-Generate synthetic data and populate the graph

-Query the graph with SPARQL

This workshop provides a hands-on environment for participants to explore how to take machine learning systems from notebooks to production. Starting from classical DevOps foundations like version control, testing, and CI/CD, participants will learn how these practices extend to the ML lifecycle, covering experiment tracking, model versioning, and reproducible project structures. Building on these foundations, participants will design and implement deployment pipelines adapted to ML workflows, and learn how to keep production systems healthy.

‍

Given the fast-moving MLOps tooling landscape, the workshop content will be periodically updated to match current state of the art and best practices.

‍

This workshop provides a hands-on environment for participants to explore how to use Large Language Models (LLMs) for a variety of tasks. From simple tasks like structured data extraction, through to more advanced topics like Retrieval-Augmented Generation (RAG). Building on these foundations participants will learn how to put together an agentic system which is able to complete any open ended tasks it is given the tools to tackle. Finally participants will learn how to evaluate the system they've built up and what considerations apply when the system is taken from prototype to production.

‍

Lastly, the workshop will cover some basics of agentic tools for software development, their benefits, drawbacks and how to use them in a way that puts safety and security front and centre.

‍

Given this is a rapidly evolving technology, the workshop content will be periodically updated to match current state of the art and best practices.
‍

‍

Together with the rapid adoption of AI, regulatory expectations, public scrutiny, and the potential for unintended harms are rising. This workshop equips participants with the conceptual foundations and practical tools needed to adopt AI systems in an ethical and responsible manner.

‍

We begin with an introduction to the ethics of AI in which we examine the concrete harms that AI systems can cause and the mechanisms through which these harms arise. Participants will also be introduced to current AI governance frameworks and learn how these can be operationalised to manage risks.

‍

The second part of the day offers a deep dive into four core ethical principles — safety, privacy, fairness, and transparency. For each, we examine how risks relating to each principle materialise, how they can be assessed, and which technical and organisational mitigations are most effective.

‍

Finally, an interactive session woven through the day follows an example use case across its full project lifecycle.  This exercise allows participants to apply governance practices, risk assessments, and mitigation strategies as they would within their own organisations.

The whole workshop can be tailored for a more technical audience (typically data scientists) or for decision makers without a technical data science background.

This workshop introduces a comprehensive framework for causal analysis, enabling participants to understand and model cause-effect relationships across domains, such as medicine (e.g., clinical trials), retail (e.g. customer engagement), economics (e.g., policy evaluation), and technology (e.g., recommendation systems).

‍

We begin by examining the limitations of predictive models and correlations, highlighting why they are often insufficient for answering “what if” questions or reasoning about interventions. To overcome these challenges and avoid common pitfalls, we present the foundations of causal reasoning, including counterfactuals, interventions, and causal graphs, and demonstrate how these concepts apply to real-world problems.

‍

The workshop is structured around two core areas: causal discovery and causal inference. Causal discovery focuses on uncovering underlying causal structures directly from data. We will explore a range of methods, with particular attention to their assumptions and limitations. Causal inference, in contrast, aims to estimate the effect of interventions or treatments on outcomes. We will cover both quasi-experimental approaches and modern machine learning techniques, along with practical strategies for evaluating their validity.

‍

Finally, two extended hands-on sessions will allow participants to apply these theoretical concepts in practice and quickly get started with causal machine learning using Python.

"AI for Decision Makers" (AI4DM) is a 5-days course for executives and decision makers. This course focuses on achieving impact and innovation with data science. It features theoretical lectures on selected applications of data science, practical lectures on leveraging data science within a business context, and a selection of exemplary use cases from the private and public sector. AI4DM helps leaders navigate fast-changing technologies and rising expectations by focusing on real-world impact. You will learn how to evaluate data and AI opportunities, build trust in analytics, and drive innovation across your organization.

‍

Who Should Attend

Ideal for executives, CDOs, digital and innovation leaders, project owners, and managers responsible for data initiatives and data-driven decisions. No technical or coding knowledge is required.

‍

What You Will Learn

* Data & AI strategy for leaders: Understand key data science, machine learning and AI concepts and their business implications.

* Opportunity evaluation: Assess feasibility, risk, cost, and value using proven frameworks.

* From concept to MVP: Explore how high-impact data products are designed and tested.

* Responsible & trustworthy AI: Address data quality, ethics, explainability, and governance.

* Leading transformation: Measure impact, manage stakeholders, and support adoption.

‍

How Can You Register

AI4DM is provided every year at ETH Zurich and at EPFL for individual professionals, and it is provided on-demand for organisations with the SDSC membership for their employees.

This workshop gives a technical introduction to the main techniques used in reinforcement learning and its applications. The topics of this course are the following:

-Introduction to Reinforcement Learning (Markov Decision Processes, Bellman Equation, Dynamic Programming, Monte Carlo methods, SARSA and Q-learning) - 2h

-Deep Q Network (DQN) - 1h

-Policy Gradient (REINFORCE algorithm) - 1h

-Actor-Critic Methods (A2C, A3C and SAC) - 2h

-Deterministic Policy Gradient algorithms (DDPG, TD3 and D4PG) - 2h

-Trust Region Policy Optimization algorithms (TRPO and PPO) - 1h

-Advanced Techniques in Reinforcement Learning (HER and RAINBOW) - 1h

-AlphaGo, AlphaGo Zero, AlphaZero and MuZero - 3h

-AlphaStar - 1h

‍

The total length of this course is 15h but it can be given in a shortened version with a selection of topics.

Bias mitigation techniques can be used to ensure fairness and impartiality in the outcomes of machine learning models, using robust data collection, preprocessing, in-processing, and post-processing methods. In this seminar, we will present the most popular techniques, and we will show some examples of their application.