Experience
Education
Bio
My research goal is to develop machine learning systems that enable computer vision to match human vision capabilities in terms of generality, robustness, and flexibility. I am convinced that this requires machine learning models to integrate compositional representations, generative modeling, and recurrent reasoning. Among other applications, I used these principles to develop methods that learn efficiently from limited supervision, generalize in previously unseen domains, and systematically analyze and overcome bias in machine learning systems.