Abstract: Upcoming experiments such as the Simons Observatory, DESI and LSST probe the universe with extremely high resolution. This upcoming data provides us with great opportunities for fundamental physics, such as probing the initial conditions of the universe. However, the vastness and extreme complexity of this interrelated data requires new methods to make sense of it. I will describe two ways forward, one based on theoretical understanding and one based on computation. In the former, I show how CMB and galaxy data can be combined in a new way to give unprecedentedly tight constraints on aspects of primordial physics. In the latter, I describe a step towards performing precision cosmology with machine learning, in a way that builds on the powerful physical and statistical methodology that has led to the success of observational cosmology.