We present recent results on the use of machine learning in studying various aspects of Calabi-Yau manifolds relevant to string theory. In particular, we discuss calculations of topological data associated with Calabi-Yau hypersurfaces in toric varieties, the construction of reflexive polytopes using genetic algorithms, and generating triangulations of reflexive polytopes providing fibrations of Calabi-Yau manifolds with reinforcement learning relevant in type IIA and F-theory. We also present techniques for computing physically normalized Yukawa couplings in heterotic string compactifications using the standard embedding, including machine learning Ricci flat metrics on Calabi-Yau manifolds.