In order to obtain the optimal design scheme of a foundation pile that meets the reliability requirements, the construction cost of foundation piles should be reduced as much as possible under the condition of effectively controlling the settlement of foundation piles. In this paper, the foundation pile of China Zun Mansion is taken as an example, and the inherent uncertainty in ultimate resistance of soils is considered. The non-dominated sorting genetic algorithm-II (NSGA-II) is used, which includes the non-dominated solution selection sorting and crowing distance comparison for individuals in a population. In this algorithm, the pile diameter and pile length are regarded as optimal design variables, the targeted reliability index of the bearing capacity for foundation pile is taken as constraint condition, and the minimum engineering cost and the lowest settlement is taken as an objective function. The Pareto optimal solution set is obtained, and a few shortcomings are resolved, such as the poor optimization performance and slow speed of traditional multi-objective optimization methods in the absence of experience. Moreover, each solution in the Pareto optimal solution set is weighted by using the TOPSIS method based on the entropy weight theory, the best scheme is chosen with the largest relative closeness. The results show that the chosen scheme is slightly better than the original design scheme in terms of construction cost and pile settlement, which demonstrates that the proposed optimal design method for foundation piles is feasible.