research themes | Matt L. Wiemann

I am interested in the role of machine learning in science, and the role of physics to build more capable artificial intelligence. I approach these problems through the three main avenues below.

~/latent_dynamical_systems.md

Latent dynamical systems

Learning compact, physically meaningful representation spaces in which complex systems evolve, making their dynamics easier to learn, easier to constrain (if one desires), and generally more robust to systematics.

~/causal_world_models.md

Causal world models

Models that capture the generative and causal structure of physical systems, supporting intervention, counterfactual reasoning, and reliable extrapolation rather than naive pattern-matching.

~/ai_agents_for_discovery.md

AI agents for scientific discovery

Autonomous and semi-autonomous systems that combine learned dynamics, physical priors, and reasoning to form hypotheses, design experiments, and accelerate scientific discovery.