Trace is building the data marketplace for physical AI, focusing on transforming how work is done in the real world. The role involves owning the spatial perception layer of the data pipeline, ensuring reliable representations from raw sensor data, which is critical for the performance of the platform.
Responsibilities:
- 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
- Train and/or fine-tune models for pose estimation and semantic understanding of multi-modal data
- Diagnose and fix the failures that actually show up in the field – drift, calibration drift, sensor misalignment, degraded tracking, weak reconstructions, noisy data
- Define the ground-truth and benchmarking methodology we use to know whether the spatial layer is actually getting better
- Decide where we need custom perception work versus where off-the-shelf components are good enough
- Work closely with the rest of engineering and with Trace Labs (our applied research arm) to feed reliable spatial outputs into downstream annotation, evaluation, and product workflows