UK remote
Research Scientist - 3D Diffusion
About this role
SpAItial is pioneering the next generation of World Models, pushing the boundaries of generative AI, computer vision, and the simulation of reality. We are moving beyond 2D pixels to build models that natively understand the physics and geometry of our world. Our mission is to redefine how industries, from robotics and AR/VR to gaming and cinema, generate and interact with physically-grounded 3D environments. We're looking for bold, innovative individuals driven by a passion for pushing the boundaries of generative 3D AI.
You should thrive in an environment where creativity meets technical challenge and be fearless in tackling the hardest problems in 3D world modeling. Our team is built on a foundation of dedication and a shared commitment to excellence, so we value people who take immense pride in their work and place the collective goals of the team above personal ambition. As a part of SpAItial, you'll be at the forefront of building World Models that bridge generative AI and the physical world.
If you're ready to make an impact, embrace the unknown, and collaborate with a talented group of visionaries, we want to hear from you. We're seeking a Research Scientist focused on 3D diffusion. You will lead research to design, build, train, evaluate, and optimize diffusion-based generative models that produce high-quality 3D content from images, video, and other inputs, with an emphasis on world-scale scenes that are spatially consistent and physically grounded.
Responsibilities - Design and develop diffusion-based methods for 3D generation from images, video, and other inputs. - Build, train, optimize, and evaluate 3D diffusion models, including research on architectures, losses, and sampling strategies. - Apply and adapt cutting-edge image and video diffusion backbones (e.g., Stable Diffusion, FLUX, WAN, or comparable systems) to 3D generation. - Implement and experiment with state-of-the-art 3D representations including point clouds, meshes, and 3D Gaussian Splatting.
- Develop training pipelines and loss functions that improve geometry accuracy, visual fidelity, and spatiotemporal consistency. - Collaborate with researchers to integrate physics-aware priors and world model capabilities into diffusion systems. - Analyze model performance, debug failure cases, and iterate rapidly to improve quality and robustness. Key Qualifications: - PhD in computer science, computer vision, graphics, machine learning, or a related field.