Seminars 2026/27

The seminars are open to anyone from the university, students are encouraged to attend. This page will be updated as more speakers are confirmed.

Next Seminar: October 7

Dr Grant Stevens, University of Bristol

When AI Pipelines Meet Scientific Reality: Assumptions, Constraints and Complexity

Machine learning methods are often developed under assumptions that do not hold for real-world scientific data, which may be noisy, incomplete, heterogeneous, constrained by physical processes, or difficult to label. Drawing on my work applying AI and machine learning methods to astrophysical problems, this talk examines the practical consequences of that mismatch, including degraded performance, methodological failure, and approaches that become impractical for domain experts. While the examples are drawn from astrophysics, the methodological issues they expose extend well beyond the field and apply broadly across scientific ML.

Semester 1

Semester 2