Intelligent tutoring systems and online classes have the potential to revolutionize education. Realizing this potential requires tackling a large number of challenges that can be framed as machine learning problems. We will first provide a survey of several machine learning problems in education, such as modeling a student's thought process as she solves a problem, constructing the atoms of knowledge, and automated problem design. We will then discuss cognitive modeling and instructional policy construction in more depth, and describe state-of-the-art methods as well as ongoing challenges. Throughout the tutorial we will highlight where student learning results in opportunities for new algorithmic and theoretical advances in machine learning.