I will present a numerical method for variational problems under b-convexity constraints, which generalize convexity constraints and are motivated by the principal-agent problem in economics. Convexity of such optimization problems requires conditions on b identified by Figalli, Kim and McCann. For a class of b which satisfy these conditions and are tractable numerically, we apply an iterated projection approach, namely Dykstra's algorithm, to projection problems. This is a joint work with Guillaume Carlier (Université Paris-Dauphine).