Trace Labs

Senior Computer Vision Engineer

United States · Senior · Full-time

Sponsorship not specified$150k-$300kDetected 57 days ago

RESTComputer VisionRoboticsSensorsAR/VRResearchCommunication

About the role

If calibration, localization, and trajectory recovery are unreliable, everything downstream - hand and pose annotation, object understanding, scene labeling, policy training - gets worse.
Doing this well makes the entire output of Trace better, and our customers feel it immediately.
The work spans calibration, localization, mapping, pose estimation, and the failure modes that show up when you run perception systems against real-world data at scale.

Own camera and multi-sensor calibration across our capture rigs, including intrinsics, extrinsics, and time synchronization
Build, evaluate, and improve SLAM, VIO, and mapping pipelines that recover aligned 6-DoF trajectories from real-world captures
Pragmatism about when to use off-the-shelf components, when to build custom, and when to push a problem back to the sensor or capture side
We build the infrastructure that makes it possible to capture and transform real-world data from humans performing physical work, and turn it into training data for robotics systems, embodied AI models, and other AI systems that operate outside the browser and in the physical world.

Strong experience in at least one of: SLAM, visual odometry, VIO, mapping, or localization
A track record of shipping perception systems on real hardware, in real-world environments - robotics, autonomy, AR/VR, drones, or other embodied / sensor-heavy systems
Comfort reasoning across software, sensors, calibration, and data quality, not just models in isolation
Experience with reconstruction, SfM, pose graph optimization, or bundle adjustment

But today, progress is constrained by a fundamental limitation: there is no scalable way to collect high-quality, real-world training data.
Frontier robotics models are trained on orders of magnitude less data than language models because there is no equivalent of an "internet of robotics data."
Trace exists to change that.
Our platform is designed to support many data formats, capture workflows, and customer needs over time.
What we capture today is only the starting point.
If you want to be an early hire at a company helping define how robots learn to work, keep reading.

This is early.
Your work will materially shape the product, systems, and direction of the company.
We are building core infrastructure that enables many future products and use cases as physical AI evolves.
The specific sensor stack we capture on today will evolve over time, so we are looking for someone who is comfortable reasoning across software, sensors, and data quality rather than someone tied to a particular pipeline.
Our platform is designed to support many data formats, capture workflows, and customer needs over time.

This listing is sourced directly from Trace Labs's careers page and normalized into a canonical job model.