Uncertainty is inherent when making decisions. Robust optimization and distributionally robust optimization have become widely adopted tools for building solutions that remain effective under uncertainty, but some challenges have always limited their practical functionality: high computational effort in terms of both memory and runtime, conservative models of the uncertainty leading to suboptimal solutions, and complex formulations that require advance knowledge of convex analysis. In this talk, I will address these challenges by discussing new data-driven frameworks for optimization under uncertainty. In particular, I will focus on reducing data dimensionality to achieve computational speedups, and building decision-focused uncertainty sets to reduce conservatism.