Differentiable simulation
Methods for optimizing through elastic bodies, cloth, fluids, articulated systems, contact, friction, and other dynamics central to graphics and robotics.
Why This Workshop
Differentiable physics is moving from a specialized simulation technique into a shared substrate for graphics, physical AI, reconstruction, optimization, and fabrication.
Physical representations
Converting visual and generative representations, from images and meshes to neural fields and generated assets, into simulation-ready geometry, constraints, and material parameters for computation.
Scalable systems
GPU programming, robust solvers, parallel simulation environments, and differentiable frameworks that make optimization and learning loops feasible at useful scale.
Various applications
Applications in system identification, control, generative modeling, digital fabrication, garment and scene understanding, and embodied interaction with the physical world.
Speakers
The keynote and lightning speakers are researchers working across avatar systems, robotics, GPU simulation, geometry processing, physical AI, generative content creation, and fabrication.
Keynote
Tuur Stuyck
Senior Manager, Robotics · NVIDIA
cloth simulation · robotics · avatars · differentiable dynamics
Leads simulation work spanning robotics, high-performance graphics, machine learning, and virtual avatars.
Keynote
Ming Lin
Professor · University of Maryland
physics-based interaction · robotics · haptics · simulation
A leading voice in physically based modeling, real-time interaction, robotics, haptics, and virtual environments.
Keynote
Miles Macklin
Senior Director · NVIDIA
GPU simulation · Warp · Isaac · differentiable programming
Builds robust GPU frameworks for simulation and differentiable programming used in graphics and robotics workflows.
Keynote
Denis Zorin
Silver Professor · NYU Courant
geometry processing · simulation · scientific computing
Works on geometric modeling, multiresolution surface representations, fluid and solid simulation, and computer graphics.
Lightning
Guying Lin
PhD Student · CMU
physics-augmented 3D generation · vision · neural fields
Studies physics-augmented reconstruction and generation, including physically plausible text-to-3D scene creation.
Lightning
Yunuo Chen
Postdoctoral Scholar · UCLA
physics-aware generation · simulation · visual computing
Works on physics-based simulation and generative models that preserve spatial, temporal, and physical structure.
Lightning
Kemeng Huang
PhD Candidate · HKU
GPU IPC · Libuipc · contact · high-performance simulation
Develops GPU-accelerated Incremental Potential Contact methods for robust, accurate, differentiable simulation.
Lightning
Dewen Guo
Postdoc · University of Utah
solids and fluids · GPU optimization · digital fashion
Develops scalable optimization methods for solids, fluids, coupled systems, real-time fashion, and multibody dynamics.
Lightning
Ying Jiang
Postdoctoral Researcher · NVIDIA
human-AI creation · generative physical AI · embodied AI
Builds geometry- and physics-guided generative systems for 2D, 3D, and 4D content creation and physical world modeling.
Lightning
Yifei Li
PhD Candidate · MIT CSAIL
physical AI · differentiable simulation · computational design
Combines physics simulation, optimization, and learning so AI systems can reason about and act on physical dynamics.
Schedule
Monday, July 20, 2:00-5:30 PM, Room 406 AB. Keynotes include 30 minutes of talk plus 5 minutes of Q&A; lightning talks include 5 minutes of talk plus 1 minute of Q&A.
2:00-2:05 PM
Opening
Minchen Li
2:05-2:40 PM
Keynote 1
Tuur Stuyck
2:40-3:00 PM
Lightning Talks I
Guying Lin, Yunuo Chen, Kemeng Huang
3:00-3:35 PM
Keynote 2
Ming Lin
3:35-4:10 PM
Keynote 3
Miles Macklin
4:10-4:30 PM
Lightning Talks II
Dewen Guo, Ying Jiang, Yifei Li
4:30-5:05 PM
Keynote 4
Denis Zorin
5:05-5:30 PM
Freeform Discussion
Open problems, benchmarks, practical barriers, and collaboration opportunities.
Organizers
The organizing team connects simulation, graphics systems, robotics, physical AI, and differentiable computation across academia and industry.
Chenfanfu Jiang
Professor · UCLA
Yin Yang
Associate Professor · University of Utah
Tuur Stuyck
Senior Manager, Robotics · NVIDIA
Bring physics into the loop.
DPGAI is for researchers, practitioners, and students who want differentiable simulation to connect pixels, geometry, materials, policies, and fabrication into one editable physical pipeline.
Monday, July 20 · 2:00-5:30 PM · Room 406 AB
Join us for four keynotes, two lightning sessions, and a closing discussion on the algorithms, systems, benchmarks, and representations needed for differentiable physics in graphics and AI.