Our premise is simple and ambitious: in the deepest levels of the progress in AI contain mathematics. We expand the frontier of AI with mathematics, and in turn use AI to advance mathematics and frontiers in science.

That two-way exchange runs through everything we do — such as the theory of why deep networks learn, learnable operators that solve scientific equations, analysis of deep generative models and more. The themes below are not walls. most projects sit across several of them. Each links to a more detailed description, related publications and more.