String theory provides our most developed framework for unifying quantum mechanics and gravity, but connecting it to the physics of our Universe requires a detailed understanding of how ten-dimensional geometry gives rise to four-dimensional cosmology. In this talk, I will discuss recent progress toward this goal using a combination of new analytic techniques and large-scale computational tools. I will highlight our recent construction of explicit candidate de Sitter solutions, as well as new mechanisms for generating hierarchies of scales. I will further discuss how Bayesian methods let us map axion spectra across the string landscape and confront them with data from the CMB, and next-generation axion experiments.