Previously, Max researched how machine learning approaches can be used for physical sciences and looked for new particle physics interactions in neutron decay. To collect data for the analysis, he prepared and managed an experiment at the Institut Laue-Langevin, France. Also, he worked on technical AI safety research to study backdoored language models and their inner mechanisms.

The goal of his research is to work towards a safe and responsible use of AI to reduce risks and benefit society. Max's research focuses on studying large language models and their emergent capabilities with interpretability methods. During his fellowship, he will work with Prof. Clark Barrett from the Stanford Computer Science Department and the Stanford Center for AI Safety.