3 months ago
Responsibilities
- Design and optimize hand and full-body human pose estimation systems for teleoperation and motion-capture annotation.
- Develop camera calibration, hand-eye calibration, visual SLAM, scene reconstruction, and temporal sensor-alignment pipelines.
- Train and evaluate object detection, instance segmentation, and 6DoF pose-estimation models.
- Optimize real-time perception inference with TensorRT and CUDA on embedded robot platforms, including custom CUDA kernels.
- Build automated annotation and Auto QA pipelines that convert sensor data into training labels and filter low-quality data.
- Collaborate with ML engineering and data infrastructure teams to connect perception outputs with downstream VLA model training.
- Establish feedback mechanisms linking perception accuracy, data quality, and model-training outcomes.
Requirements
- 5+ years of industry experience in robot perception or computer vision.
- Strong 3D vision fundamentals, including stereo cameras, structured-light cameras, and 3D reconstruction.
- Proficiency with SLAM frameworks such as ORB-SLAM, VINS-Mono, and FastLIO, or equivalent V-SLAM development experience.
- Hands-on human pose-estimation experience using hand-joint or full-body pose methods such as MediaPipe, MANO, OpenPose, or SMPLify.
- Proficiency with deep learning training frameworks for perception model training, tuning, and evaluation.
- Experience deploying TensorRT models and optimizing real-time inference on embedded platforms such as Jetson or Horizon.
- CUDA programming fundamentals and ability to write or debug custom kernels.
- Proficiency in C++ and Python with ROS or ROS2 development experience.
- Proficiency with AI coding agents.
- Experience with 6DoF object pose estimation, 3D Gaussian Splatting, NeRF, robot manipulation or teleoperation, automated annotation pipelines, or ground-truth generation is preferred.
- Published research in perception, pose estimation, or robotics is preferred.
Benefits
- Direct involvement in producing high-quality robot training data for embodied intelligence.
- Work alongside robotics engineers and machine-learning researchers.
- Access to proprietary hardware platforms including humanoid robots, camera arrays, and data gloves.
- Fast-paced, high-autonomy 0→1 work environment.
Categories
ML EngineeringRobotics