The jamming transition in granular materials has garnered significant interest for its potential in applications such as soft robotics, energy-absorbing structures, and adaptive systems. However, conventional granular materials, primarily composed of convex particles, suffer from inherent mechanical limitations, including low tensile and bending strength in the jammed state. In this talk, I will introduce a Bayesian optimization-based design framework that enhances interlocking mechanisms in granular materials, leading to improved shear and bending performance. I will also present experimental results demonstrating the adaptability of these structures and their promising applications in reconfigurable architectures and soft robotics.