My research focuses on generative media systems, image and video foundation models, diffusion and flow models, inference optimization, 3D vision, and computer vision.
For the full publication list, see Google Scholar.
Generative Media & Foundation Models
Movie Gen: A Cast of Media Foundation Models
Movie Gen is a family of media foundation models for video generation and editing. My team contributed to the efficiency workstream, reducing generation latency from tens of minutes to seconds and supporting research and production deployments.
Imagine Flash: Accelerating Emu Diffusion Models with Backward Distillation
Imagine Flash enables high-quality real-time image generation by accelerating diffusion models. I led the technical strategy and execution of the project, which moved from research to production in approximately four months.
Bespoke Non-Stationary Solvers for Fast Sampling of Diffusion and Flow Models
Research on fast sampling methods for diffusion and flow models, improving the efficiency of generative model inference.
Published at ICML 2024.
Human Motion, Avatars & Generative Systems
Avatars Grow Legs
Generative motion system for producing smooth full-body human motion from sparse tracking inputs.
Published at CVPR 2023.
StyleAvatar
Research on stylizing animatable head avatars.
Published at WACV 2024.
3D Vision & Computer Vision
Re-ReND
Real-time rendering of neural radiance fields across devices.
Published at ICCV 2023.
VisCo Grids
Surface reconstruction using viscosity and coarea grids.
Published at NeurIPS 2022.
ASSANet
Efficient point cloud representation learning using anisotropic separable set abstraction.
Published at NeurIPS 2021.
DeepGCNs
Research exploring whether graph convolutional networks can go as deep as convolutional neural networks.
Published as an ICCV Oral in 2019.
Full Publication List
For a complete and updated list of publications, visit my Google Scholar profile.