Senior Robot Perception Engineer (SLAM & Pose Estimation)

XDOF; San Mateo (CA); On-site; USD 170,000 - 250,000; Full time

Job summary

XDOF is pioneering perception for autonomous robots in California. The Perception team shapes data collection and annotation pipelines, builds 3D pose estimation, and optimizes real-time inference on embedded platforms.

You will advance hand and full-body pose estimation, calibrate multi-camera rigs, and develop end-to-end data-to-model workflows enabling high-quality robot training data.

Qualifications

• 5+ years of industry experience in robot perception or computer vision.
• Strong 3D vision fundamentals: stereo and structured-light principles.
• Proficiency with SLAM frameworks (ORB-SLAM, VINS-Mono, FastLIO) or V-SLAM development.
• Hands-on experience with human pose estimation: hand joints or full-body.
• Proficient in deep learning training frameworks for perception model work.
• TensorRT deployment experience with real-time inference on embedded platforms.
• CUDA programming fundamentals; write or debug custom kernels.
• Proficient in C++ and Python with ROS/ROS2 development.
• Proficient with AI coding agents

Responsibilities

• Design and optimize hand pose estimation pipelines for teleoperation data.
• Build full-body pose estimation systems for motion capture ground truth.
• Research vision-based pose estimation to reduce data collection costs.
• Fuse pose estimates with robot joint data to generate training annotations.
• Design and maintain calibration pipelines for multi-camera arrays.
• Build visual SLAM / V-SLAM for real-time localization and scene reconstruction.
• Implement hand-eye calibration between cameras and end-effectors.
• Develop temporal alignment across multimodal sensors (cameras, IMU, data gloves).
• Train and iterate on perception models including 6DoF pose estimation.
• Optimize model inference using TensorRT/CUDA for embedded platforms.
• Write custom CUDA kernels for acceleration of perception tasks.
• Design evaluation metric frameworks and monitor performance vs data quality.
• Contribute to automated annotation pipelines from sensor data.
• Build Auto QA modules to filter low-quality data.
• Collaborate with ML engineers and data infra to meet training needs.
• Establish feedback linking perception accuracy to training outcomes.

Skills

5+ years experience in robot vision/CS

3D vision fundamentals

SLAM/V-SLAM familiarity

human pose estimation experience

deep learning for perception

CUDA programming fundamentals

C++ & Python with ROS/ROS2

AI coding agents

Tools

ROS/ROS2

TensorRT

CUDA

OpenPose

SMPLify

MANO

MediaPipe

Jetson

Horizon

FoundPose

Job description

XDOF is pioneering perception for autonomous robots in California. The Perception team shapes data collection and annotation pipelines, builds 3D pose estimation, and optimizes real-time inference on embedded platforms.

You will advance hand and full-body pose estimation, calibrate multi-camera rigs, and develop end-to-end data-to-model workflows enabling high-quality robot training data.