Research Scientist - 3D Diffusion

SpAItial is pioneering next‑generation World Models that combine generative AI, computer vision and physics‑based 3D simulation. As a Research Scientist focused on 3D diffusion, you will lead the design, development and optimisation of diffusion‑based generative models that create high‑quality 3D content from images, video or other inputs, targeting world‑scale, spatially consistent and physically grounded scenes. You will adapt state‑of‑the‑art image and video diffusion backbones such as Stable Diffusion, FLUX or WAN for 3D generation, implement and experiment with 3D representations including point clouds, meshes and Gaussian splatting, and build training pipelines, loss functions and sampling strategies that improve geometry accuracy, visual fidelity and spatiotemporal consistency. The role requires close collaboration with other researchers to integrate physics‑aware priors and world‑model capabilities, rigorous analysis of model performance, debugging of failure cases and rapid iteration to enhance robustness. Candidates must hold a PhD in computer science, computer vision, graphics, machine learning or a related field, have a strong publication record at top venues, and possess deep expertise in deep learning, diffusion models and large transformer architectures. Proficiency in Python and PyTorch, experience with large‑scale model training, and solid knowledge of 3D processing concepts such as camera geometry, depth reconstruction, point clouds, meshes or Gaussian splats are essential.