Students at Université de Technologie de Compiègne (UTC) are allowed to design their curriculum by selecting courses from large course pools. There are today six engineering degrees (high-level master). The assignment is performed using a heuristic designed some years ago. The number of students, courses and curricula increased over the years and this heuristic reached its limit. Time-consuming manual post-processing is required to obtain the individual planning for each student not assigned by the heuristic. We propose a MIP model to maximize the number of assigned students while managing limited resources. Optimal solutions are achieved for each instance and are strictly better than those computed by the heuristic. We investigated the impact of clique inequalities on computing times.