Human fingers possess compliant contact interfaces with predetermined compliance properties. Replicating these compliance properties in robotic limbs, particularly anthropomorphic hands, can significantly impact grasp stability and contact area. To address the limitations of homogeneous layers of soft material commonly used in anthropomorphic hands, this paper proposes optimizing differentiated structures to design soft pads, which consist of a continuous external layer (skin) coupled with an internal layer featuring voids. To achieve the desired compliance properties replicating those of human fingers, the paper introduces five distinct patterns for differentiated structures, along with their corresponding finite element analysis (FEA) models. Subsequently, by formulating an appropriate optimization function, the shapes of these patterns have been optimized to match the compliance of the human finger, and it has been demonstrated that one of the proposed and optimized pads closely approximates this target through the proposed methodology. The results affirm that this approach offers an effective solution for designing soft pads whose behavior needs to be aligned with the compliance of a specific target.