Research Scientist - 3D Diffusion

Job description

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. - Top-tier publication record at venues such as CVPR, ECCV/ICCV, NeurIPS, and SIGGRAPH. - Strong fundamentals in deep learning and generative modeling, in particular diffusion models and large transformer models. - Hands-on experience training diffusion models and working with cutting-edge image and video model stacks (e.g., Stable Diffusion, FLUX, WAN, or similar). - Solid understanding of 3D processing concepts such as camera geometry, depth, reconstruction, point clouds, meshes, or Gaussian splats. - Proficiency in Python and deep learning frameworks such as PyTorch, with experience in large-scale model training and optimization. - Ability to implement research ideas, run rigorous experiments, and ship reliable ML code. At SpAItial, we are committed to creating a diverse and inclusive workplace. We welcome applications from people of all backgrounds, experiences, and perspectives. We are an equal opportunity employer and ensure all candidates are treated fairly throughout the recruitment process.