Design and simulation of passive reflectors to be used in 5G applications are presented. An optimization environment is constructed based on genetic algorithms (GAs) and a full-wave solver to obtain compact reflectors that possess various reflection characteristics in accordance with given design specifications. In addition to their triangulated models for simulations, reflectors are represented by Bezier surfaces on the optimization side to reach smooth designs that are suitable for fabrication. A multigrid approach is further used to increase the effectiveness of the optimization trials via a systematic deformation of surfaces from coarse shaping to fine tuning. The results demonstrate the feasibility of achieving diverse reflection characteristics with relatively compact reflectors, as well as the effectiveness of the proposed optimization strategy.