Build the sensing and data systems physical intelligence is missing.

What you will do

• Build evaluation loops that expose where physical AI models fail, then turn those failures into precise collection targets.
• Train and evaluate imitation learning, vision-language-action, and multimodal policies on synchronized real-world data.
• Define the signals, labels, coverage, and benchmarks that make a dataset measurably useful instead of merely large.

What makes this a fit

You have strong machine learning or robotics fundamentals and like working across models, data, and systems. Experience with imitation learning, control, multimodal learning, or evaluation is useful, but we care most about unusually strong work and clear thinking.

What you will do

• Own ingestion, synchronization, validation, lineage, and versioning for high-volume multimodal data.
• Build quality, review, search, and delivery systems so model teams can request exact slices instead of receiving folders of files.
• Set the security, observability, reliability, and cost foundations across field devices, cloud systems, and customer environments.

What makes this a fit

You have built production data, distributed, media, or ML infrastructure and can own ambiguous systems end to end. You care equally about throughput, correctness, and making complex data easy for other engineers to use.

What you will do

• Launch and operate collection programs across farms, factories, warehouses, and other skilled work environments.
• Recruit and train operators, design practical procedures, and own safety, logistics, and quality in the field.
• Work with model, data, and capture teams so every shift produces the coverage, outcomes, and edge cases a training program needs.

What makes this a fit

You are relentlessly operational, earn trust quickly, and can move between site logistics, sensors, and data specifications without dropping details. Experience in field deployment, industrial operations, agriculture, robotics, or an early-stage startup is useful.