Molding parameters for curing composite structures are the major factors that govern the distortion or warpage in composite products. This paper presents a two-stage genetic algorithm (GA) procedure to inversely determine the optimal molding parameters that minimize residual stresses. Critical molding parameters include heating rate at each heating stage, temperature at the preservation stage, the duration of each temperature-preservation stage, and the cooling rate. The domain of each molding parameter is firstly specified to cover possible variation ranges in practical production. In our proposed two-stage GA procedure, a finite element model for multiphysics (heat transfer, curing kinetics, and thermal mechanics) simulation is first created using COMSOL® to compute the residual stresses of the composite laminated plate for a given temperature curve. The FEM model is then modulated by a improved GA with the residual stresses of the plate as the objective function. The improved GA is called in two-stages: the first stage determines a set of likelihoods of the modeling parameters around which the "optimal" parameters may reside. The 2nd stage zooms-in the areas centered by these likelihoods, which finds molding parameters that minimize the residual stresses.  A number of tests are conducted, and the results show that the proposed two-stage GA procedure yields near global optimal value.