Creating functional midsoles for shoes is a challenging task that involves considering different aspects such as stability, comfort, manufacturability, and aesthetics. No single approach exists to design a midsole that meets all these objectives effectively. Therefore, this study aims to introduce a multidisciplinary optimization method to develop custom shoe midsole structures. Our approach involves utilizing generative methods to generate diverse structures and leveraging swarm intelligence to search for optimal designs. Without loss of generality, we use tetrahedral mesh generation to create midsole structures because tetrahedral structures are renowned for their exceptional strength. To enhance the swarm’s exploration of the design space and discover more local optima, we developed a swarm behavior that promotes diversity. Furthermore, we created a quantitative measurement tool to evaluate various objectives. In order to test the effectiveness of our generative approach, we analyzed the midsoles generated from our design exploration that performed the best and the worst in relation to each objective. Our findings revealed a substantial difference between them, with scores differing by two to four times. Additionally, when compared to other lattice structures, the tetrahedral midsole structure created by our method demonstrated superior compliance with the foot and better redistribution of plantar stress. The multidisciplinary optimization technique we have proposed is a valuable resource for engineers and designers in the footwear industry, allowing them to develop high-performance midsole structures that meet the needs of both consumers and athletes. Furthermore, this method can be applied to optimize other complex structures in various industries, such as civil, automotive, and aerospace engineering.