Aerodynamic shape optimization has become of primary importance for the aerospace industry over the last years. Most of the method developed so far have been shown to be either computationally very expensive, or to have low dimensional search space. In this work we present how Geodesic Convolutional Neural Networks can be used as a surrogate model for a computational fluid dynamics solver. The trained model is shown to be capable of efficiently exploring a high dimensional shape space. We apply this method to the optimization of a fixed wing drone, the eBee Classic.