The team builds computer vision software that turns stereo camera captures into precise, traceable pile measurements.
You will be responsible for the core perception pipeline, moving work from research into production and making sure the measurements hold up in messy real-world conditions.
Responsibilities
Create and maintain pipelines for stereo reconstruction, depth estimation, segmentation and 3D volume estimation.
Integrate and assess vision models, including stereo networks and YOLO-style detectors.
Refine point cloud cleanup, ground plane fitting and overall reconstruction quality.
Check the assumptions behind camera calibration and find the causes of failure using field sensor data.
Build confidence measures and test measurement accuracy as prototypes move toward production.
Requirements
You bring at least 4 years of experience in computer vision or perception ML engineering.
You are solid in Python and PyTorch, with experience using OpenCV and Open3D.
You have worked hands on with depth estimation, point clouds, 3D geometry, calibration or reconstruction, including stereo vision.
You have shipped vision systems built on real sensor data, ideally with stereo perception or volume reconstruction.
You can work through sensor data problems and judge end-to-end perception quality.
Experience with CUDA, robotics perception, SLAM, photogrammetry or industrial inspection is beneficial.
Conditions
Base salary runs from $180,000 to $245,000 per year.
Visa sponsorship is offered.
The role is on-site in San Francisco, California, United States.