Project Hydra | Senior Software Engineer - Robot Data Collection

Team: Data Annotation

Location: Santa Clara, CA

Commitment: Full Time

Workplace Type: onsite

Salary:

What You’ll Do

Design, build, and maintain the software behind our data collection stations — teleoperation stack, multi-camera recording, and session tooling.
Deploy and upgrade that stack across the fleet without stopping collection; keep workstations, containers, and networking healthy.
Own the data pipeline end to end: automated validation, conversion into our dataset format, upload, metadata, and versioning.
Build the QA tooling that catches dropped frames, stream desync, bad calibration, and mislabeled episodes before data reaches training.
Direct a team of teleoperators — define task protocols and SOPs, onboard them on new tasks and embodiments, review session quality, and unblock them daily.
Own fleet reliability: calibration and firmware pipelines, sensor and end-effector integration, root-cause analysis of recurring faults, and vendor escalation.
Bring up new stations and new embodiments; report throughput, yield, and data health to the research team.

What we need to see

• MS in CS, Robotics, EE/ME, or a related field (or BS with equivalent hands-on experience).
• 3+ years building software for real robot systems — not simulation only.
• A system you designed and shipped that other people depended on daily, and that you kept running.
• Strong Python and Linux; Docker, git, CI, and SSH-based fleet administration.
• Robot middleware and real-time multi-sensor recording — ROS 2 or equivalent vendor SDKs.
• Experience supporting a data collection or teleoperation operation: you have built the tools operators depend on and owned throughput and quality numbers.
• Comfortable around robot hardware — arms or humanoids, dexterous hands, RGB-D cameras, and the calibration that keeps them honest.
• Able to lead operators without formal authority: you write clear SOPs, train people, and communicate status without being asked.

Ways to stand out from the crowd

• Humanoid or bimanual platforms (Unitree, Fourier, ALOHA/ARX-class rigs) and dexterous hands.
• XR teleoperation: Quest / PICO / Vision Pro, motion-capture gloves, whole-body retargeting.
• Imitation-learning data tooling — LeRobot, HDF5/Zarr episode stores, dataset versioning, S3 at TB–PB scale.
• Built internal dashboards or tooling for a production line or high-throughput lab.
• Open-source contributions to robotics data or teleoperation tooling.
#L1-CC1