Short Trainings & Workshops

This course offers a comprehensive introduction to the field of data science from a technical standpoint. Participants will gain a high-level understanding of how AI algorithms can learn from labeled data, focusing on supervised learning. The course will delve into the broader landscape of AI techniques, with special emphasis on the latest advancements in Large Language Models. The final segment of the presentation is dedicated to exploring real-world applications of AI.

This presentation provides a comprehensive introduction to Bayesian Optimization, a powerful technique for efficiently finding optimal solutions in expensive-to-evaluate functions by building probabilistic models and intelligently selecting where to sample next. We'll discuss how to use BO in robust optimization where an optimum that is resistant to perturbations in the input parameters is desired. Additionally, we'll discuss the use of BO in formulation optimization where both components and their proportions are jointly optimized.

This workshop provides an in-depth exploration of Large Language Models (LLMs), Prompt Engineering, and Retrieval Augmented Generation (RAG) systems. It introduces the fundamentals of LLMs, focusing on their capabilities and limitations, followed by practical techniques for crafting effective prompts. The course emphasizes the role of RAG in enhancing the accuracy and relevance of LLM-generated content by retrieving external data. Agentic LLM applications will also be introduced.Through use cases from the Swiss Data Science Center (SDSC), participants learn about the architecture, tools, and performance metrics of RAG systems, with interactive discussions encouraged throughout the workshop. A short hands-on session in Renku, will also provide a practical experience with a toy RAG (https://renkulab.io/projects/lamagioviale/rag-system-handson).

This workshop provides an introduction to recommender systems, focusing on the classical and most commonly employed methods (Content Based Filtering and Collaborative Filtering). The workshop includes presentations of business use cases where such systems have been implemented and succesfully deployed.

The first part of the workshop will introduce the different types of uncertainty, like aleatoric and epistemic ones, and define the typical uncertainty metrics. The second part will detail several techniques to define and estimate uncertainty in machine learning projects: Bootstrapping, Ensemble methods, Bayesian approaches and Quantile regression. The third part of the workshop is dedicated to a more recent and promissing technique: Conformal Prediction. The last part is a hands-on session in python to experiment with these diffent techniques, with simple datasets.

In this course we will guide the audience through the challenges of framing a data science project that can be tackled using data science techniques. Share previous experience from public and private sector. Provide hands-on tutorial on how to fill a data science canvas.

This introduction aims to position the topic of natural language processing within the wider scope of artificial intelligence applications. The foundations and state of the art of the discipline will be introduced, together with its core tasks and applications. Concrete examples of natural language processing techniques powering Swiss organizations will be presented together with final remarks concerning limitations and perspectives of the field.

This course is a high-level introduction to Generative AI techniques, with the focus on recent developments of large language models (LLMs), to demystify these techniques and discuss practical business applications.Techniques involving images, sound and tabular data, as well as multi-modal systems will be discussed.The course will also introduce basic concepts and best practices in prompt engineering.The last part of this course is dedicated to a hands-on session where participants will work with LLMs on specific tasks (summarization, generation, translation...).