With the growth of the electric and hybrid vehicle market, the optimization of electric machines is inevitable. These machines are optimally designed to respect a drive cycle composed of a considerable number of operating points. By considering saturation phenomena and all different losses in the studied machine the optimization process then becomes very time-consuming. To overcome this limitation, very fast electromagnetic models should be considered. To this end, a Gaussian Process (GP) metamodel is proposed and used to predict the electromagnetic performances and losses of a wound field synchronous machine, which is a permanent magnet-free machine solution for electric mobility. Optimal currents for armature and excitation windings are computed over the entire drive cycle using loss minimization strategy.