research

I am interested in the mechanisms of learning and reasoning in artificial and biological brains. In particular, I am driven by the observation that certain abilities, notably compositional generalization and efficient multimodal integration, are natural for humans but fickle for modern AI systems. I wish to a) investigate the conditions under which these abilities emerge for neural networks, and then use these insights to b) design more robust, interpretable, and sample-efficient generative frameworks that can reason and operate across multiple data modalities. Additionally, I am also broadly interested in diffusion models, neuroscience, representation learning, and interesting problems regarding intelligence in general.