An optimization process is used to redesign blades of a high-pressure compressor. An artificial neural network (ANN) method is coupled to Navier-Stokes solvers and is applied to three different redesigns. A new rotor blade of a transonic compressor is designed by modifying thick, stacking line and inlet angle using a 3D approach, with a significant efficiency improvement at the design point. The off-design behavior of this new compressor is also checked afterwards, which shows that the whole performance of the inlet stage is improved over a wide range of mass flow. The losses are reduced, proving the good performance of the optimum. The whole results indicate that the optimization method can find improved design and can be integrated in a design procedure.