Data Labeler $30,000 USD/year Pay is set based on global value, not the local market.
Most roles = hourly rate x 40 hrs x 50 weeks
Description
If accuracy matters more to you than speed, this position is designed for your strengths. The labels you produce become the training foundation for AI systems serving thousands of students daily. Precise behavioral annotation makes the product more intelligent. Inconsistent labeling teaches the model incorrect patterns.
LearnWith.AI develops AI-driven learning tools by combining learning science, data analytics, and domain expertise. This position transforms unstructured student session recordings into reliable, rubric-aligned labels the engineering team depends on. You will review recorded student sessions, detect critical behavioral moments, and follow explicit protocols to categorize what occurred and its timing. You will also evaluate LLM-generated pre-annotations, correct errors, and record edge cases to help engineers refine the system.
This is not freelance-style, disconnected annotation work. It is a consistent workflow within one product area, featuring direct quality feedback, calibration against reference standards, and advancement tied to precision and reliability. If you value transparent expectations, quantifiable quality metrics, and contributions that directly influence model outcomes, we should connect.
What you will be doing
- Label student session recordings by detecting, categorizing, and time-marking behavioral events according to a comprehensive rubric
- Evaluate and refine LLM pre-annotations by eliminating incorrect labels, inserting overlooked events, and sharpening timestamp accuracy
- Document clear reasoning for ambiguous decisions, including rubric citations and the logic behind your classification
- Record edge cases and clarification requests for unclear scenarios and maintain an annotation log with session details
- Participate in calibration sessions, integrate QA feedback, and adopt rubric revisions to enhance accuracy consistently
What you will NOT be doing
- Develop AI models, conduct experiments, or perform research into student behavior patterns
- Redesign the annotation rubric or modify category definitions according to personal judgment
- Prioritize throughput over accuracy, consistency, or timestamp exactness
- Handle sporadic, disconnected tasks spanning unrelated fields with no continuity or feedback mechanism
Key responsibilities
This role ensures that student session recordings are transformed into ≥95%-accurate, temporally precise labeled datasets that dependably indicate when model performance advances or declines.
Candidate requirements
- A minimum of 1 year of experience in data annotation, content moderation, QA evaluation, or comparable rubric-based review roles
- Excellent English reading comprehension and the capacity to adhere to detailed written guidelines without deviation
- Capacity to maintain concentration and precision for 4–6 hours of video-focused work daily
- Skill in identifying nuanced visual and on-screen behavioral signals and categorizing them uniformly across multiple sessions
- Solid written communication abilities for documenting edge cases, reasoning, and clarification inquiries
- Dependable internet connection sufficient for video streaming
- Confidence reviewing, correcting, and enhancing AI/LLM-produced annotations
Applying for a role? Here’s what to expect.
Crossover's skill assessment process combines innovative AI power with decades of human research, to take the guesswork, human bias, and pointless filters out of recruiting high-performing teams.
Chat-style
screening interview.
Cognitive 
aptitude test.
Prove real-world 
job skills.
Interview with the hiring manager.
Pass
proctored test.
Accept job offer.
Frequently asked questions
Reviews
Meet some people who've landed similar jobs
Recruitment sucks. So we’re fixing it.
The Olympics of work
It’s super hard to qualify—extreme quality standards ensure every single team member is at the top of their game.
Premium pay for premium talent
Over 50% of new hires double or triple their previous pay. Why? Because that’s what the best person in the world is worth.
Shortlist by skills, not bias
We don’t care where you went to school, what color your hair is, or whether we can pronounce your name. Just prove you’ve got the skills.
Join the world's largest community of 
AI-first remote workers.