Research Scientist - 3D Diffusion
Core
Design, build, train, and optimize diffusion-based generative models to produce high-quality, physically-grounded 3D content from images and video for world-scale scenes.
Role type
Research Scientist (3D Diffusion)
Builds
Production 3D generative models and world models
Domain
Generative AI, Computer Vision, 3D Graphics
Deliverable
production ML models
Required skills
Diffusion models, Large transformer models, 3D processing (camera geometry, depth, reconstruction, point clouds, meshes, Gaussian splats), Python, PyTorch, Large-scale model training, Research implementation
Preferred skills
Physics-aware priors, Spatiotemporal consistency optimization
Technologies
Stable Diffusion, FLUX, WAN, PyTorch
Responsibilities
Design diffusion methods for 3D generation from images/video; Build and optimize 3D diffusion models; Adapt image/video diffusion backbones to 3D; Implement 3D representations (point clouds, meshes, Gaussian Splatting); Develop training pipelines and loss functions; Collaborate on physics-aware world model integration; Analyze performance and debug failure cases
Seniority
Senior, hands-on IC
