Staff Software Engineer (Computer Vision)
Posted Updated
- As a Staff Software Engineer for Computer Vision, you will develop and deploy across all of Array’s 3D processing capabilities, including our photogrammetry and radar reconstruction pipelines
- The work spans computer vision algorithms and production engineering: you should be equally comfortable reading a paper about unfamiliar algorithms, and shipping new capabilities as production-quality software
- The position will work closely with 3D reconstruction scientists, radar algorithms engineers, and product engineers
- Productionize and continuously improve Array’s photogrammetry and radar image formation pipelines using industry best practices
- Step into individual pipeline stages — bundle adjustment, dense matching, point cloud generation, geometric refinement, geo-referencing — to improve accuracy, robustness, and throughput
- Develop and optimize database schemas and storage solutions for managing large-scale 3D geospatial data
- Contribute to broader software engineering efforts across the company, including data infrastructure, analytics, and systems to deliver our satellite data to customers
- Strong foundation in geometric computer vision, spanning camera modeling, epipolar geometry, and multi-view reconstruction pipelines (stereo, SfM, MVS, bundle adjustment)
- Excellent communicator, with the ability to make high-quality technical decisions in a startup environment
- 5+ years of software engineering experience, with a substantial portion focused on computer vision
- Strong proficiency in Python and C++, experience with cloud platforms (AWS, GCP, Azure) and containerization technologies (e.g. Docker, Kubernetes)
- Experience with geospatial concepts, remote sensing data, or robotics/autonomous vehicle sensor data
- Familiarity with digital signal processing, radar algorithms, or SAR concepts
- Familiarity with deep learning approaches to 3D vision (e.g. NeRF, 3D Gaussian Splatting, learned feature matching, or depth estimation)
- Experience with C++ or CUDA for performance optimization
- Ability to build visualization tools for geospatial or 3D data (Cesium, Three.js, or similar)
- We will conduct interviews via Google Meet with the potential for an onsite visit; the typical process takes around 3-5 weeks to complete from start to finish