This study explores wireless power transfer (WPT) systems for public electric vehicle charging, focusing on optimising the transmitter design to enhance interoperability across various receiver coil geometries and alignment conditions. Due to the complex non-linear relationships inherent to WPT systems, traditional optimisation methods are computationally expensive. Therefore, this study proposes an approach using artificial neural networks (ANNs) trained on finite element method (FEM) data to develop a surrogate model of the WPT system. This model is integrated into a blackbox optimisation solver, enabling faster identification of improved transmitter designs. The proposed method achieves computational speeds 6,000 times faster than traditional FEM simulations, with post-validation on the final solutions verifying prediction errors below 0.6%. The results demonstrate a significant acceleration in the optimisation process, establishing this method as an effective framework for developing practical WPT systems for public charging applications.