AI Engineer, Robot Data and Evaluation

Roboligent; Round Rock (TX); On-site; USD 110,000 - 160,000; Full time

Job summary

Roboligent, an Austin-based robotics company, is building Robin, a force-controlled fast mobile manipulator for warehouse work. You will own the data loop that turns field data into training-ready material, and you will design evaluation protocols to quantify improvements on real robots in customer environments.

You are the accountable owner of data collection tooling and the pipeline, ensuring fast, reliable data generation by field engineers and operators.

Qualifications

• 3+ years of engineering experience with data from real robots.
• Direct experience with imitation learning or robot policy learning in a role where real-world performance was your problem.
• Clear technical writing — provide evaluation protocol, data-quality document, or experiment write-up.
• Real experience developing with AI coding agents as part of your workflow.
• Willing to be in the lab with the hardware — this role does not work remotely.
• Degrees are useful signal, not a requirement.

Responsibilities

• Own the data pipeline: ingestion, versioning, time sync, conventions, automated quality checks before data reaches a training run.
• Own evaluation: a real-robot evaluation protocol, regression tracking, failure taxonomies.
• Own collection tooling and operator enablement: make field engineers and customer operators productive at generating usable episodes.
• Own data-driven model work: fine-tunes and ablations that decide where collection effort goes.
• Own the contract with the fleet: what gets captured, what leaves the customer's building, in what shape.

Skills

Imitation learning

Robot policy learning

Technical writing

AI coding agents

Robotics with real hardware

Engineering experience

Lab presence

Job description

Roboligent is an Austin-based robotics company building Robin — a force-controlled bimanual mobile manipulator for warehouse and industrial work. Our patent-pending actuators give robots human-like sensitivity, so they work alongside people instead of behind cages. Robin is deployed with real customers today, on real production floors.

The engineering team is small and the scope is focused. Work ships to real robots in real customer environments within days, not quarters.

Role Description

Most vision-language-action models get their capability from pre-training on enormous, expensive datasets. We use a few hundred well-chosen demonstrations per task to reach deployment-grade success rates.

Nobody currently owns this data loop end to end. You do: the pipeline that turns raw teleoperation into training-ready data, the evaluation harness that tells us whether a change helped, and the connection between the two.

You are the AI team's accountable owner for the data loop. You are not the person who collects the data — our field engineers and operators do that — but part of your job is making them fast at it.

What you'll own

• The data pipeline — ingestion, versioning, time sync, consistent conventions, automated quality checks before data reaches a training run
• Evaluation — a real-robot evaluation protocol the team can trust, regression tracking, failure taxonomies that tell us how a policy failed
• Collection tooling and operator enablement — making field engineers and customer operators productive at generating usable episodes
• Data-driven model work — the fine-tunes and ablations that decide where collection effort goes
• The contract with the fleet — what gets captured, what leaves the customer's building, in what shape

What this role is not

• You are not the data collector — that belongs to the field team. You own what gets collected and the tooling that makes it cheap.
• You do not own model architecture — that sits with the founder and AI team today; this role influences it through evidence, not design authority.
• You are not a labeling or annotation manager.

Qualifications

• 3+ years of engineering experience, including meaningful time working with data from real robots
• Direct experience with imitation learning or robot policy learning, in a role where the model's real-world performance was your problem
• Clear technical writing — we will ask to see an evaluation protocol, data-quality document, or experiment write-up you wrote
• Real experience developing with AI coding agents as part of your workflow
• Willing to be in the lab with the hardware — this role does not work remotely
• Degrees are useful signal, not a requirement

Location and logistics

On-site in Austin, Texas. Not remote, not hybrid. Occasional domestic travel to customer sites during proof-of-concept windows; not a travel-heavy role.

Must be authorized to work in the U.S. without sponsorship, now or in the future.

Process

About three weeks end to end, with feedback after every stage: an intro call with our CEO, two writing samples (an evaluation protocol or data-quality document, and an agent configuration file), a technical screen with the AI team, a paid 3-hour working session with a real robot dataset, final conversations, and references.