Job Description
At XDOF, we’re at an inflection point. Frontier labs are racing to build general-purpose robots, and high-quality training data is the bottleneck. We’re building the foundation behind the foundation models – the data collection systems, operational capability, exabyte-scale data warehouse, and software toolchain – to help our partners drive the field forward.
We're looking for a full-stack engineer who can work across the entire stack, from hardware, sensors, and data pipelines through to algorithm deployment, and who will independently own the end-to-end implementation of embodied AI capabilities on robots.
Responsibilities
Maintain and iterate on the embodied AI data collection system, covering multimodal sensor synchronization, data pipelines, storage, and downstream annotation toolchains.
Own the integration, calibration, and routine maintenance of camera systems (ZED / RealSense / industrial cameras, etc.) and robotic arms, keeping equipment stable and operational.
Independently own end-to-end deployment of imitation learning / DAgger / teleoperation algorithms on robots, including data collection, model training, inference deployment, and on-site tuning.
Drive data quality assurance mechanisms: timestamp synchronization, coordinate frame alignment, anomalous data detection, and rapid diagnosis and repair of collection failures.
Define and maintain robot-side software interfaces and communication protocols (ROS2 / DDS, etc.) to enable fast replication of collection stations and deployment stations across platforms and sites.
Support on-site data collection and algorithm deployment, produce technical documentation, and continuously improve collection efficiency and model performance.
Requirements
Bachelor's degree or above in Robotics, Automation, Computer Science, or a related field, with 3+ years of relevant experience.
Strong programming skills in Python / C++; familiar with system-level development and debugging in Linux environments.
Expert in ROS / ROS2, including node communication mechanisms (Topics, Services, Actions), parameter management, and the Launch system.
Familiar with camera system integration and calibration (intrinsics, extrinsics, multi-camera synchronization); hands-on experience with ZED / RealSense or similar RGB-D / stereo cameras.
Familiar with robotic arm control interfaces (e.g., UR / Franka / xArm or in-house arms), with an understanding of the driver layer, kinematics, and high-level API calls.
Familiar with common robot hardware interfaces (CAN, EtherCAT, RS485, USB) and sensor integration (IMU, depth cameras, force/torque sensors, etc.).
Experience training and deploying deep learning models; proficient in PyTorch and able to independently handle the full workflow from data preparation and training to inference and deployment.
Proficient in using AI coding tools (e.g., Claude Code, Cursor) in daily development, with a demonstrated ability to significantly boost engineering productivity through them.
Nice to Have
Complete project experience in embodied AI / imitation learning data collection and algorithm deployment.
Familiarity with training and deployment details of mainstream imitation learning algorithms such as Diffusion Policy, ACT, and VLA.
Familiarity with robot data formats such as SVO2, ROS bag, HDF5, and LeRobot datasets.
Familiarity with multi-device synchronization solutions such as camera hardware triggering and PTP time synchronization.
Experience integrating devices such as VIVE Trackers, OptiTrack, and motion capture gloves.
Experience with video encoding/decoding (H.264 / H.265 / Jetson hardware codecs).
Fluent in written English, able to read research papers and open-source project documentation directly.
What We Offer
End-to-end ownership of engineering deployment for embodied AI capabilities, with full autonomy from data to algorithms to deployment.
A competitive compensation package.
A flat, open engineering culture with genuine involvement in technical decisions.
Well-equipped hardware lab resources supporting rapid prototyping and iteration.
Flexible work arrangements, plus ongoing learning and internal tech-sharing programs.