Many modern data objects, including distributions, networks, trees, shapes, and points on manifolds, do not naturally lie in a Euclidean space. My research develops statistical learning methods for these complex objects using the distances between them, with a focus on regression, Bayesian inference, and meta-learning.