This paper presents a hybrid (Genetic Algorithms, GA and Adjoint Method, AM) tool for the optimum design of aerodynamic shapes. CFD techniques for unstructured grids are used during both optimization phases. The best solution computed after a few generations through the GA costitutes a good starting shape for the second, AM-based optimization phase. This is enriched in design variables through a newly proposed technique before proceeding to the AM phase. Different parallelization techniques are proposed for each phase, using the PVM message passing protocol. During the genetic evolution the concurrent evaluations of candidate solutions is used whereas the domain decomposition technique is employed during the AM-based optimization.