Teams at Niantic Spatial, Inc.

Develop algorithms and production systems for world-scale geospatial registration, 3D geometry, correspondence, mapping, and learned 3D reasoning. Build datasets, pipelines, benchmarks, and evaluation methods using imagery, point clouds, LiDAR, GPS, and geographic data. Apply computer vision, machine learning, transformers, and geometric optimization to improve a living world map, while collaborating across research, engineering, and product teams and communicating results through technical documentation and presentations.

Conduct research and develop algorithms for 3D reconstruction, computer vision, spatial AI, neural scene representations, Gaussian splatting, and generative models. Publish at leading conferences, build large-scale experiments, prototype rapidly, and transition research into customer-facing production systems. Collaborate with researchers and engineers across computer vision, machine learning, graphics, robotics, and geospatial modeling. The role requires strong Python and PyTorch skills, a 3D vision research record, top-tier publications, and evidence of shipping research-driven products.

Train and evaluate PyTorch computer vision models for visual localization, including feature matching, image retrieval, pose regression, and scene-coordinate regression. Build large-scale image and 3D data pipelines, run experiments on GPU infrastructure, maintain localization benchmarks, and support model export, inference optimization, deployment, and monitoring. Collaborate with research and testing teams to improve production model performance.