This study explores possible designofan aeroacoustic flame deflector forarocket launch pad. The design objectives are 1) minimization of the overall sound pressure level near the payload fairing, 2) minimization of the time-averaged maximum pressure on the flame-deflector surface, and 3) minimization of the shape difference from a flat plate inclined at 45 deg. The acoustic wave characteristics associated with deflector shapes are identified by large-eddy simulations. To overcome difficulties of required computational time, the following are adopted: 1) a high-order scheme that reduces the computational cost of large-eddy simulations, 2) amulti-objective evolutionary algorithm for efficient parallelization, and 3) large-scale parallelization on the Japanese supercomputer "K". Total computational time for optimization is approximately 350 h with 6500 processors of the K computer. The analysis of nondominated (Pareto-optimal) solutions reveals a tradeoff relation and correlation among the objective functions. In the result, there appears a well-balanced solution that significantly reduces the overall sound pressure level. The shape difference is relatively minor, with a small bump located somewhat upstream of the impinging region. The result suggests that the local angle of the inclined deflector near the impinging region plays an important role for the reduction of overall sound pressure level near the rocket fairing.