AI Engineer, Computer Vision — Localization & Production

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

Job summary

Niantic Spatial is hiring an AI Engineer, Computer Vision, to contribute to deep learning models for their localization system. Youll train and iterate visual models, working with researchers and engineers to improve feature extractors, matchers, and pose regressors across a full training-to-production loop.

Youll gain exposure to data pipelines, evaluation, and deployment, with emphasis on correctness and scalable experimentation.

Qualifications

• Experience training deep learning models for computer vision tasks using PyTorch.
• Solid understanding of visual localization, image retrieval, pose estimation, and 3D geometry.
• Experience handling large image/3D datasets and data preprocessing.
• Familiarity with GPU-based training and analyzing training runs to identify bottlenecks.
• Strong Python programming skills and clean, reproducible code.
• Bachelor's or master's degree in CS/CV/Robotics or equivalent practical experience.

Responsibilities

• Train, evaluate, and iterate visual localization models in PyTorch.
• Build data pipelines for imagery, poses, and 3D point clouds.
• Run scalable experiments across GPU clusters and ensure reproducibility.
• Extend the evaluation suite for pose accuracy and latency.
• Support production deployment from checkpoint to serving.
• Bridge research and product to translate field failures into improvements.

Skills

PyTorch

Python

Deep learning

Computer vision

GPU training

Data preprocessing

Education

CS/CV/Robotics degree

Tools

CUDA

TensorRT

ONNX

Job description

Niantic Spatial is hiring an AI Engineer, Computer Vision, to contribute to deep learning models for their localization system. Youll train and iterate visual models, working with researchers and engineers to improve feature extractors, matchers, and pose regressors across a full training-to-production loop.

Youll gain exposure to data pipelines, evaluation, and deployment, with emphasis on correctness and scalable experimentation.