Our Research
Research in the DIGIT Lab aims to answer this overarching question: How can AI design better materials, structures, and machines? We look beyond model performance and efficiency to focus on four key challenges:
- Complexity. Advances in manufacturing technology have unlocked significant design complexity, directly resulting in dramatic improvements in functional capabilities. Yet, this potential is often constrained by the limits of human-led design. The gap between what we can build and what we can design highlights a fundamental challenge in modern engineering. Our work confronts this challenge by exploring the question: How can AI unlock design complexities that are fundamentally inaccessible to traditional methods?
- Creativity. Human creativity is constrained by individual knowledge and experience, with outcomes often highly uncertain. Meanwhile, today’s AI excels at generalizing within the boundaries of existing data and knowledge instead of generating genuinely new ideas. To transform AI from a tool of interpolation into an engine for innovation, we must answer a critical question: How can AI enable or accelerate the discovery of “out-of-the-box” design solutions?
- Trustworthiness. For AI-generated designs to be adopted in critical, high-stakes applications, engineers must be able to trust them. Many powerful AI models operate as “black boxes,” providing solutions without clear explanations or uncertainty quantification. This lack of transparency creates a significant barrier to adoption. Our research aims to answer the question: How can we develop AI-driven design methodologies that engineers can trust?
- Insights. The ultimate promise of AI in design is not just to create better products, but to make us better designers. We aim to develop methods that can decode the complex patterns learned by AI into human-understandable knowledge, such as new design rules or unknown structure-property relationships. This goal drives our research to answer a fundamental question: How can AI generate new knowledge that expands human understanding and guides design?
Guided by these fundamental questions, we develop computational methodologies for design ideation, generative design, and design for X (“X” can represent manufacturing, sustainability, reliability, and beyond) across diverse engineering domains.
For more information, you can explore our paper collection or check out our open-source code.