<p>Mounted protrusively on the train roof, the pantograph serves as an essential current-collection component for high-speed trains. At high operating speeds, the aerodynamic lift and drag acting on the pantograph vary considerably, which directly affects the performance of the pantograph–catenary system. In this study, a surrogate model-based approach is adopted to establish the correlation between the geometric design variables of the recessed platform and aerodynamic responses, followed by a sensitivity evaluation of the design variables. The results reveal that the sensitivity ranking of design variables differs across various aerodynamic responses. The platform depth (<i>z</i>) exerts a comparable influence on the standard deviation of pantograph lift force (<i>σ</i>_plf) and base drag (A_bd). The length of the upstream platform section (<i>x</i>_<i>1</i>) presents the strongest positive effect on the mean pantograph lift force (A_plf); this variable affects A_plf and <i>σ</i>plf with similar magnitudes but opposite trends. Taking <i>σ</i>_plf and Abd as the optimization objectives, the Non-dominated Sorting Genetic Algorithm II (NSGA-II) is employed for the multi-objective optimization of the recessed platform, and a set of Pareto optimal solutions for the design variables is obtained. Several optimal solutions are randomly selected for validation. Generally, the predictions from the surrogate model show good agreement with the Computational Fluid Dynamics (CFD) results for A_plf, <i>σ</i>_plf, and A_bd, with the largest discrepancy observed in <i>σ</i>plf. Although minor deviations exist in a few optimized cases, the maximum relative error is less than 13%. Overall, the optimized platform designs achieve remarkable aerodynamic improvements, particularly in reducing lift force fluctuation and base drag.</p>