Computer Vision Seminar Series: Dr. Minghao Guo, "Building Scientific World Simulators at Scale."
Nov 20, 2026 12:30 pm
2405 Siebel Center for Computer Science or Zoom

- Sponsor
- Siebel School of Computing and Data Science
- Speaker
- Dr. Minghao Guo
- Contact
- Hao-Yu Hsu
- haoyuyh3@illinois.edu
- Originating Calendar
- Siebel School Speakers Calendar
- Abstract: Generative foundation models have reshaped modern design workflows. Yet extending their impact from code and images to the physical world of science and engineering imposes a stricter requirement: generated designs must stay physically consistent under real dynamics and constraints. I argue that the next leap will come from a unified, learnable simulation model that spans many physical regimes at once: a scientific world simulator. Rather than only trading off accuracy against speed within one or few physics domains, as numerical simulators do, it generalizes across designs and physical regimes. This generality comes from pretraining on multi-physics data, which extends the classical idea of precomputation, keeping it fast and faithful enough to drive the iterative loop of computational design. As such, it forms a new layer of abstraction above physics simulation and real-world experiment, exploring candidate designs broadly and quickly before deferring to exact solvers and experiments for verification. In this talk, I will trace this arc through my recent work on scalable physics pretraining and differentiable projection modules, building toward scientific world simulators that advance computational design and scientific discovery, from engineered shapes to new proteins.Speaker Bio.: Minghao Guo (https://www.minghaoguo.com/) earned his PhD at MIT, where he was advised by Prof. Wojciech Matusik at CSAIL. He is currently a Postdoctoral at MIT, working with Profs. Kaiming He and Wojciech Matusik, and a Visiting Scholar at Stanford University. He received his M.Phil. from the Chinese University of Hong Kong and his bachelor’s degree from Tsinghua University. His research lies at the intersection of computer graphics, generative modeling, and scientific discovery. He develops generative design systems that translate scientific and engineering intent into physically valid, manufacturable, and reliable designs. His work has received the Outstanding Systems Paper Award at RSS 2026 and the Best Paper Award at the ICLR 2026 Workshop on Foundation Models for Science. He is also a recipient of the MathWorks Fellowship and the Meshy AI Fellowship, and was a finalist for the Roblox Fellowship.
Zoom Password: csillinois