Machine Learning Engineer: Perception AnalyticsWe are looking for a perception engineer to build features for our customers' analytics. This isn't just a data science job. This is developing fundamental perception features to drive customer-facing platforms. What are the real-world challenges in measuring dig productivity from lidar/camera data? Can we modify existing perception systems to provide useful metrics for job sites? How do you measure safety over an entire site given perception systems? How do you re-identify the same objects across dozens of camera views on a site several city blocks wide, through dust and changing light?
What You'll Do:- Extend, modify and adapt existing perception signals for analytics needs
- Work with customers to define and scope what is possible with advanced perception systems on-site
- Deploy models and analytics directly to fleets of machines
What We're Looking For:- MSc or advanced degree in Computer Science, Robotics, or a related field
- 4+ years of professional experience shipping perception systems to production (ideally on robotic or other embedded platforms), with strong hands-on experience in a modeling framework (e.g. PyTorch)
- Proficient in Python and comfortable reading and writing at least one systems language (e.g. C++, Rust)
- Hands-on experience incorporating raw sensor data (camera, lidar, IMU) into learned pipelines
- Solid grounding in 3D geometry, sensor calibration (intrinsics/extrinsics), coordinate transforms, and camera/image re-projection algorithms
- Strong data analysis skills across statistical characterization of sensor data, corner-case and anomaly discovery, and evaluation design
Ways to Stand Out:One or more of the following:
- Prior experience working with customers to understand perception data and analytics
- Experience with the Rust programming language
- Published work in top-tier venues such as ICRA, IROS, CVPR, ECCV, ICCV, CoRL, or RSS
Other special aspects of the role:- Based in the Bay Area with the ability to be onsite at our SF office on a daily/weekly basis
