Research Scientist - 3D Reconstruction & Novel View Synthesis

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 seeking a Research Scientist to push the state of the art in 3D reconstruction. You should have command of the latest scene representations, such as Gaussian splats and radiance fields, and of reconstruction methods across the full range, from per-scene optimization to feed-forward models. We are looking for someone who knows the field well enough to choose the right representation and method for a problem, and then improve on it. This is a senior, hands-on research role for someone who has already done reconstruction research across more than one representation.

Responsibilities

• Advance the state of the art in 3D reconstruction, from research ideas to working methods.
• Work across the full range of reconstruction problems (geometry, appearance, materials, lighting and dynamics, from sparse or incomplete observations), and novel view synthesis, including neural rendering and feed-forward view generators.
• Design losses, priors, and optimization strategies that improve fidelity, robustness, and efficiency.
• Build rigorous evaluations on public and internal benchmarks, and use them to drive decisions.
• Collaborate with research and engineering colleagues to bring successful methods into production systems.
• Key qualifications
• PhD in computer vision, graphics, or a related field; publications at top venues (CVPR, ICCV, ECCV, SIGGRAPH, NeurIPS).
• Deep, hands-on knowledge of modern scene representations, for example Gaussian splats and radiance fields, and the ability to weigh their trade-offs.
• Research experience across the reconstruction spectrum, from per-scene optimization to feed-forward models.
• Deep understanding of multi-view geometry, camera models, and rendering.
• Strong experience with deep learning frameworks.