About Me
I am the Chief Scientist at Viggle AI, where I lead research on generative video models, from new model capabilities to production.
Previously, I was a founding member and Staff Researcher at Waabi, where I developed Waabi World, a generative closed-loop sensor simulator for self-driving. Before that, I was a Research Scientist at Uber ATG R&D. I am also completing my Ph.D. at the University of Toronto, advised by Prof. Raquel Urtasun.
My research focuses on generative models and world models: video generation, few-step distillation, and 4D reconstruction and simulation. I am passionate about finding simple, elegant solutions to challenging problems.
Beyond research, I am an advocate of Unix philosophy, a Linux enthusiast, and the author of the bestselling PyTorch book.
There should be one – and preferably only one – obvious way to do it. Although that way may not be obvious at first.
Research
Experience
Leading research on generative video models.
- Viggle-Animate: character replacement from a single edited frame, distilled to three forward passes
- Meridian: geometry-guided video model with independent control of camera motion and scene time
- Viggle-Turbo: 5x faster few-step distillation of Qwen-Image-2.1
Working closely with Prof. Raquel Urtasun and Prof. Shenlong Wang on 3D simulation.
- 3D Reconstruction with large-scale weakly-labelled data in the wild
- Photorealistic Image Simulation with Geometry-Aware Composition for Self-Driving
Working closely with Ming Liang and Bin Yang in ATG R&D for 3D Perception tasks.
- Depth Completion: Densify LiDAR with image guidance, SOTA in KITTI
- 3D Perception: 3D detection, tracking with multi-sensor. New SOTA in KITTI
- Map structure learning with graph neural network. New SOTA in Argoverse motion forecasting
Working on Medical Imaging in Machine Intelligence Group led by Dr. Xian-Sheng Hua
- Developed 3D R-CNN for CT/MRI, faster and more accurate than radiologist.
- Explored weakly-supervised learning with limited labels and active learning for efficient labelling.
