I am broadly interested in developing and studying machine learning models that can reason about the rich structure of the physical world and its combinatorial complexity. This includes topics in graph representation learning, object-centric learning, and causal representation learning.