Labelbox's expert network for coding, STEM, finance, legal and computer-vision AI training.
Outlier review
Scale AI's expert platform where graduates and specialists write prompts, build rubrics and rate AI answers.
Outlier is run by Scale AI and connects subject-matter experts with AI companies that need human feedback to improve large language models. It says it has paid more than $500 million to experts and onboarded over 900,000 graduates, master’s students and PhDs.
What the work looks like
The three most common tasks are:
- Writing challenging prompts: you think of a hard question in your field that an AI model gets wrong, then write the correct answer.
- Creating grading rubrics that define what a good answer looks like.
- Rating and ranking answers from different models.
How to get accepted
Onboarding has four steps: create a profile, import and review your skills, verify your identity, then pass skill screenings. General onboarding takes 30 to 90 minutes, and each project has its own extra onboarding. Apply in the domains where you are strongest, because screenings decide which projects you are matched with.
Pros and cons
Pros
- No minimum hours, work whenever you want
- Weekly payouts
- Free access to paid AI models while working
- Large range of domains, from generalist to PhD-level STEM
Cons
- Project availability can stop suddenly
- Onboarding and skill screenings take 30 to 90 minutes before any paid work
What you need to apply
- Valid government ID and a mobile phone from your country of residence
- Up-to-date resume
- LinkedIn profile showing education and work history
- Usually at least undergraduate-level knowledge in your field, plus good written English
Ready to apply?
This link goes to the official website. Disclosure.
Similar platforms
Remote AI training contracts with some of the highest published rates, from generalist writing to expert coding.
Toloka's AI training platform with entry-level and expert projects in 90+ domains and 20+ languages.