Speaker: Dr. Mengdi Wang

https://wang-mengdi.github.io/

Topic: From Physics to Intelligence: High-Performance Simulation Across Scales

Abstract: Physical simulation provides a computational window into the dynamics of the real world—from the evolution of bubbles and foams to large-scale fluid motion and deformable fabrics. Yet faithfully modeling such phenomena remains challenging because they span vastly different spatial and temporal scales. This talk presents a series of GPU-accelerated simulation frameworks that address these multiscale challenges through new physical representations, adaptive algorithms, and high-performance system design. These efforts include a particle-based model for thin-film and bubble dynamics, an adaptive hybrid particle–grid framework for large-scale fluid simulation, and a real-time knit-fabric system that captures yarn-level deformation for interactive design and rendering. Together, they demonstrate how physically grounded models and GPU computing can work hand in hand to achieve both accuracy and efficiency in complex physical systems.Looking ahead, the talk will briefly discuss how such simulation techniques can connect with modern generative AI models—toward intelligent, physics-aware digital worlds.

Biography: Mengdi Wang is a final-year Ph.D. candidate in the School of Interactive Computing at Georgia Tech, advised by Prof. Bo Zhu. His research centers on high-performance GPU simulation and numerical algorithms for modeling multiscale physical phenomena. He has published multiple papers at ACM SIGGRAPH and the Journal of Computational Physics, covering topics such as adaptive hybrid particle–grid methods, real-time knit deformation, and geometric interface tracking. His ongoing work extends these techniques toward integrating physical simulation with AI-driven modeling and generation.