AI Engineer - Computer Vision for Localization

Niantic Spatial, Inc.; San Francisco (CA); Hybrid; USD 166,000 - 221,000; Full time

Job summary

Niantic Spatial, Inc. is seeking an AI Engineer, Computer Vision to advance the deep learning models powering our localization system.

You will train and iterate visual models—feature extractors, matchers, pose regressors, and retrieval models—that turn a single image into a precise position in the world. This full‑loop role offers collaboration with researchers and engineers, exposure to data, training, evaluation, and production deployment in a hybrid San Francisco office.

Qualifications

• Experience training deep learning models for computer vision in PyTorch.
• Foundational understanding of visual localization, learned feature matching, image retrieval, pose estimation, SLAM, or 3D geometry.
• Comfort working with large image or 3D datasets and data preprocessing.
• Familiarity with GPU-based training and reading training runs to identify bottlenecks.
• Strong Python skills and write reproducible code.
• Bachelor's or Master's in CS/CV/Robotics or equivalent experience.

Responsibilities

• Train and evaluate visual localization models in PyTorch, including feature extraction, matching, image retrieval, and pose regression.
• Build high-throughput data pipelines for raw imagery, poses, and 3D point clouds into training datasets.
• Run scalable training experiments across GPU environments with reproducibility in mind.
• Extend the evaluation benchmark for localization across lighting and season changes.
• Support production deployment: export, inference optimization, and monitoring of live models.
• Bridge Research and Product by translating field failures into training data and benchmark improvements.

Skills

PyTorch

Deep Learning

Computer Vision

Python

GPU Training

Education

Bachelor's or Master’s in CS/CV/Robotics

Tools

TensorRT

ONNX

CoreML

Job description

Niantic Spatial, Inc. is seeking an AI Engineer, Computer Vision to advance the deep learning models powering our localization system.

You will train and iterate visual models—feature extractors, matchers, pose regressors, and retrieval models—that turn a single image into a precise position in the world. This full‑loop role offers collaboration with researchers and engineers, exposure to data, training, evaluation, and production deployment in a hybrid San Francisco office.