In this paper we propose a course recommendation system based on historical grades of students in college. Our model will be able to recommend available courses in sites such as: Coursera, Udacity, Edx, etc. To do so, probabilistic topic models are used as follows. On one hand, Latent Dirichlet Allocation (LDA) topic model infers topics from content given in a college course syllabus. On the other hand, topics are also extracted from a massive online open course (MOOC) syllabus. These two sets of topics and grading information are matched using a content based recommendation system so as to recommend relevant online courses to students. Preliminary results show suitability of our approach.