Website:
joinshift.com
Job details:
About shift
shift pays everyday workers to record short videos of the tasks they already do (in kitchens, workshops, job sites, bakeries, and warehouses) to help train the next generation of AI and robotics. We've paid out over $15M to 25,000+ workers across 15+ countries.
The role
Every recording that enters shift's training set passes through a quality gate. You are that gate.
You'll run quality assurance on egocentric work sessions submitted by contributors across 15+ countries - deciding what ships and what goes back, and handling the escalations that come in when a decision is disputed. A large share of the first pass is automated, so your job is to audit the model's calls, catch what it gets wrong in both directions, give the contributor a reason they can act on, and turn every miss into a sharper rule.
This is high-volume, high-judgement work with real consequences on both sides. A wrong accept puts bad data in front of a model. A wrong reject costs a real person their pay and, eventually, their place on the platform. We need people who take both seriously and hold a consistent quality bar under queue pressure.
What you'll do
- Review recorded work sessions against our acceptance criteria: valid task and domain, genuine workplace environment, camera wearer performing the task, clear and stable framing, complete task progression
- Apply the rejection taxonomy consistently, and write feedback specific enough that the contributor knows exactly what to change next time
What we're looking for
- No prior experience in QA required, but a sound judgement under ambiguity and requires critical thinking
- Demonstrable consistency: you make the same call on the same evidence week after week, and can explain why
- Comfort applying a detailed violation taxonomy to messy, ambiguous, real-world footage
- Clear written English - your rejection reasons are read by contributors across multiple countries
Bonus
- Experience with egocentric or first-person video for robotics or embodied AI
- Experience auditing model-generated decisions or labels and feeding errors back to an ML team
- Hands-on with annotation or review tooling - CVAT, Labelbox, Encord, V7, Label Studio, or an equivalent internal platform
- Experience authoring or maintaining SOPs, rubrics, or calibration sets
- Background supporting distributed or global contributor workforces
Click on Apply to know more.