Hardware and software co-design is one of our future research directions. Form and function are co-dependent in robotics as they are in nature. For optimum functionality, each robotic design requires a different controller and vice versa. Thanks to the improved ability of new machine learning methods that can handle a large set of parameters, we can combine the design and controller parameter spaces and generate better robotic systems. Faster and more reliable 3D-printing methods and materials allow us to realize new robotic designs within the machine learning loop. In this research direction, we focus on the co-design of robotic grippers, which have a dynamic workspace with rich interaction configurations.