Arturs Berzins

I am a researcher at Iliad working on deep learning theory for AI safety. I study how loss-landscape geometry relates to training dynamics and learned structure, using ReLU networks as a tractable setting. I also mentor at the Iliad Fellowship

Previously, I was a postdoc at the Institute for Machine Learning at JKU Linz with Johannes Brandstetter and Sepp Hochreiter, where I worked on machine learning for geometry and simulation. I received a PhD in mathematics from the University of Oslo with a thesis on Neural Representations in Geometry. I was hosted by the Geometry group at SINTEF as a fellow of GRAPES.

I received my BSc and MSc in mechanical engineering from RWTH Aachen, specializing in simulation technology and applying machine learning to geometry-heavy engineering tasks. My broader research interests include geometry in machine learning, interpretability, deep learning theory, and machine learning for science and engineering.