Tunable stiffness structures allow robots to transition between a soft state which enables adaptive and gentle interactions and a stiff state which enables strength and precision. One effective and versatile way to achieve adaptability is through jamming, a structural phenomenon in which a structure composed of smaller elements (unjammed state) behaves as a cohesive unit (jammed state) when external forces are applied. Jamming structures based on different element geometries (e.g. fibers, particles, and layers) have been shown to deliver application-specific tunability, with each jamming modality offering advantages under different loading regimes. Prior research has investigated the systematic identification of high-performing dimensional parameters in jamming structures with pre-determined element morphologies. However, the inverse-design of freeform element geometries based on a desired bulk mechanical behavior, remains unexplored, limiting the potential of jamming structures for complex loading conditions and geometries. We propose a method which (1) takes a bulk geometry under a specific loading condition in which mechanical tunability is desired, (2) considers this cohesive structure as the “jammed” state, and (3) introduces cuts throughout the structure, creating the “unjammed” state by dividing it into smaller elements, with the ultimate goal of achieving maximum change in stiffness without compromising structural integrity. We have developed a computational design framework that deploys this method which relies on finite element simulations and gradient-based optimization. Accordingly, a cost function that considers the stiffness change and strain energy distribution is proposed. The introduced method and framework expands the design space of jamming structures, enabling the development of tunable stiffness components with arbitrary geometries under arbitrary loading conditions for soft robots or adaptive mechanical systems.