E-Learning has blossomed into an active research domain and industry, using computation to enhance educational content, pedagogy and administrative support for teachers and institutions. These efforts to complement traditional education, however, are dwarfed in scale by shadow education: a system of private tutoring that students with limited financial resources cannot access. This system, in turn, creates disparities that threaten social cohesion. In this project, we investigate whether computation can provide key benefits of private tutoring at substantially lower costs. For mathematics tutoring, we investigate whether computation can help students by (a) guiding students step-by-step through a particular problem, and (b) by automatically generating problems in the same domain that are matched to students’ observed skill-levels. This project is being conducted in partnership with Vidya Next, which offers tablet-based private tutoring.