Multi-objective design optimization was applied to the impeller and volute of a centrifugal pump using surrogate-based optimization techniques and three-dimensional Reynolds-averaged Navier–Stokes (RANS) analysis. The objective functions used to improve the hydraulic performance and operating stability of the pump were the hydraulic efficiency at the design condition and the flow rate at which the maximum volute pressure recovery coefficient occurs. Three design variables were selected based on the results of a sensitivity analysis: the blade outlet angle, the constants in determining the impeller outlet width, and the cross-sectional area of the volute. Using response surface approximation (RSA), surrogate models were constructed for the objective functions based on numerical results at experimental points obtained by Latin hypercube sampling (LHS). The representative Pareto-optimal solutions obtained by the multi-objective genetic algorithm (MOGA) show enhanced objective function values compared to the baseline design. The results of unsteady calculation show that the flow instability of the centrifugal pump was successfully suppressed by the optimization.