Operate a handheld UMI gripper in a San Francisco lab, performing scripted manipulation tasks — picking up objects, opening containers, simple household actions — to generate training data for robot learning models. No robotics or AI background is required, but the work is on-site, physically repetitive, and judged on protocol fidelity rather than creativity.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

What the work actually involves

You'll spend shifts in a lab holding a UMI (Universal Manipulation Interface) gripper — a handheld rig with cameras that records your hand motion as demonstration data. A written protocol tells you exactly what to do: pick up this mug by the handle, open this drawer, pour from this container, repeat across a set of objects and starting positions. Between takes you reset the scene to its documented starting state, then check that the recording captured cleanly before moving on. Most of the day is repetition with small controlled variations, because that variation is what the model learns from.

The other half of the job is quality control on your own output. Compromised recordings — occluded cameras, dropped frames, a gripper that drifted out of frame, an object that moved during reset — are worse than no recording, because they enter the dataset as noise. You're expected to notice, discard, and flag rather than push through to hit a count. Hardware quirks and ambiguous instructions get escalated to the project lead in writing, not improvised around.

What the screen looks for

micro1's screening is AI-led and short. It won't test robotics knowledge — it will test whether you can be relied on to follow a fifteen-step protocol identically for six hours, whether you can describe concrete past work that required manual precision and punctuality, and whether you'll admit to a mistake instead of hiding it. Expect follow-ups that push on specifics: which shift, how many hours, what went wrong, what you did. Vague self-praise about attention to detail is the main way candidates lose the screen.

Logistics

On-site only, San Francisco. This is not a remote or async role; there is no work-from-home option because the hardware and scene setups live in the lab.
Contractor engagement, scheduled in shift blocks. Hours are often structured around student availability, but you must be physically present for full sessions.
Observed pay band $10–20/hr, stated as seen on the platform and not guaranteed.
Expect sustained standing, bending, and fine hand work; physical stamina is a genuine requirement rather than boilerplate.

College Student – AI Robotics Trainer (No Experience Required)

What the work actually involves

What the screen looks for

Logistics

Interview prep — 10 questions