Summary
The first measures of the EU’s AI Act went into effect February, 2025. Fines for non-compliance can be as high as 35 million EUR or 7% of a company’s global, annual revenue, whichever is higher.
This workshop will start with an overview of the AI Act and other relevant regulation geared towards practicing data scientists and those studying data science. In the second half, we show how sound data science practices already cover most of the technical requirements of the AI Act using an example from private health insurance.
Syllabus
- AI regulation technical requirement
- Exploratory data analysis, with focus data quality and bias detection
- Feature creation from a mix of numerical and text tabular data
- Predictive model training and selection practices
- Testing and documentation
Instructors
Paul Larsen is a freelance data scientist with many years of experience in both large corporations and startups. He has worked primarily in highly regulated financial service industries, and has recently been teaching courses and workshops on the EU’s AI Act.
Attendee equipment prerequisites
A dedicated GitHub repo will be shared in advance of the workshop.
Exercises will be done in groups, so participants should either bring a laptop or pre-arrange a group that has a laptop.
Familiarity with python and Jupyter notebooks is required for the practical exercises.
Helpful but not strictly necessary are experience with
- exploratory data analysis
