Robot Infrastructure Engineer - Collection Tooling

Job description

Job responsibilities

• Build teleoperation and operator tools for leader arms, VR, and other interfaces, including calibration, action mapping, latency, and reliable recording.
• Implement episode workflows for task instructions, start and stop, resets, annotations, success labels, and human takeover, keeping a clean boundary between policy actions and human corrections.
• Build collection quality checks for frame drops, timestamp alignment, missing signals, calibration validity, and incomplete episodes, with feedback operators can act on immediately.
• Build collection specifications and adapters that validate data delivered by external partners against task requirements, schemas, calibration, and provenance.
• Keep raw recordings as the source of truth and make normalization and export to training formats reproducible.
• Show collection progress and coverage against the task so the team collects the missing variations and hard cases, not more of the same.
• Test everything on real robots and document setup and troubleshooting so engineers, operators, and partners can run collection without you in the room.

Job requirements

• Built data collection, teleoperation, sensor recording, or human-in-the-loop tooling used on real robots or autonomous systems, and validated it on hardware.
• Strong Python skills and software engineering fundamentals, including APIs, tests, versioned schemas, and debugging data pipelines.
• Experience with synchronized video, robot state, actions, or other multimodal time-series data.
• Working knowledge of coordinate frames, sensor calibration, data provenance, and how collection errors show up in trained models.
• Experience building tools for other people, watching them fail in real use, and simplifying without hiding data-quality problems.
• The ability to turn a task description into an executable collection workflow together with learning, hardware, and operations colleagues.