The prevalence of online education systems provides opportunities to deliver personalized learning at scale. Educational systems need to assess students so that they can provide better curricula tailored to each student's unique needs. Since there is a limited amount of time for quizzing a student, we need to test each student using those questions that capture the most information about their level of understanding of various concepts. In this paper, we formally pose the problem and present multiple approaches for learning a quizzing policy to determine a personalized sequence of questions for each student that best predicts their knowledge state. We first introduce simple heuristics including random selection and an uncertainty sampling approach inspired by an active learning framework. We then develop a reinforcement learning (RL) approach for designing a quizzing policy. Using simulations of students' knowledge states, we provide initial evidence that an RL-based approach can improve over simple heuristics. We further demonstrate the effectiveness of our approaches using a real-world dataset consisting of over 1.5 million examples of students' answers to mathematics questions from Eedi, an online educational platform. [For the full proceedings, see ED615472.]