A method is presented for finding the optimal autoclave temperature history for pre-impregnated thermoset composites based on their failure performance. A coupled finite element model which incorporates a thermo-chemical and incremental elastic analysis is used to predict the residual stress distribution at the edge of a thick composite beam. The optimal autoclave temperature is sought by formulating an objective function and an associated set of constraints. The objective is designed to maximize the minimum failure load of the manufactured beam amongst a set of load cases, while the constraints are imposed to ensure that the composite is uniformly cured and does not sustain temperature damage during the manufacturing process. A novel semi-analytic gradient evaluation technique is developed which incorporates elements of the adjoint and direct sensitivity methods. Initial results show that significant improvements in the failure load can be achieved for certain beam constructions.