How it works
Simple to start.
Serious about quality.
Four steps from application to first paid project. You may already be an expert — in your language, your culture, or your professional field — and not even realize it. Here’s how the platform works and what to expect inside.
The process
How to Start Earning
1
Apply
Submit your application and choose a track — domain expert or generalist. No CV screening, no recruiter calls. You’re assessed on what you can actually do. Your name, email, and domain selection. That’s it to start.
2
Assessment
Complete a domain-specific evaluation — a real sample of the work. This screens for genuine expertise, analytical clarity, and attention to detail. Not something you can cram for. 45–90 minutes, at your own pace.
3
Projects
Once approved, you’re inside a purpose-built contributor dashboard where projects are matched to your skill level. Browse available tasks, start when you’re ready, stop when you’re done. No fixed hours, no check-ins, no minimum commitment. As you complete more work and build your quality score, you become eligible for higher-complexity — and higher-paying — projects across more verticals.
4
Get paid
Payment issued directly after each completed project. No batch cycles, no waiting. Withdraw whenever you choose. Rates are competitive and vary by domain, task complexity, and region.
The platform
Inside the platform.
This is what you’ll actually see after you’re approved. Not a gig marketplace — a structured contributor environment built for serious AI training work.
🗂️
Your task queue
Projects matched to your domain and skill level appear in a personal dashboard. You choose what to work on and when. No quotas, no minimum hours, no pressure to be available at a specific time.
📚
Built-in training
New to a task type? Training modules are built into the platform. Get proficient on new formats, expand your skills, and unlock access to more work — without leaving the platform. This is how generalists get started and grow.
📈
Progress over time
Your contributor score grows with each completed project. Higher scores unlock more complex, better-paid work — including emerging verticals like agentic AI and robotics. The platform rewards quality and consistency, not just volume.
💳
Payment & history
Every completed task shows what you earned. Payment is issued after completion — no batch cycles, no waiting. Your full earnings history, project record, and quality scores all live in one place.
Not sure you qualify?
You may be an expert and not realize it.
If you use apps on your phone, speak a native language fluently, have cultural knowledge, or use search engines regularly — there are tasks here built for exactly your kind of expertise. The training built into the platform will get you up to speed on the rest.
Task types
The work, in practice.
The job isn’t one thing. Depending on the project and domain, you’ll be doing one or more of these task types — each requiring a different kind of expert judgement.
01 · Evaluate
Rate and rank AI responses
Read a question and two or more AI-generated responses. Choose which is better and explain your reasoning — grounded in domain expertise, not preference. The most common task type across all domains.
Common in: all language model domains, agentic AI
02 · Annotate
Write the ideal response
Given a prompt, write the response the AI should have produced. Your answer becomes training data that directly shapes how the model reasons in your domain. Higher cognitive load — and higher pay.
Common in: law, medicine, coding, mathematics
03 · Verify
Fact-check and correct
Review a specific AI claim for factual accuracy. Identify what’s wrong, explain why, and provide the correct information with citations where relevant. Requires deep domain knowledge to catch subtle errors.
Common in: medicine, STEM, finance, law
04 · Red-team
Find failure modes
Deliberately probe AI systems for weaknesses — unsafe outputs, reasoning failures, boundary violations, or domain-specific blind spots. Adversarial by design. Among the highest-value work on the platform.
Common in: agentic AI, robotics, medical imaging
05 · Multimodal review
Evaluate beyond text
Review AI outputs that involve images, diagrams, audio, or video alongside text. Assess whether the model correctly interpreted the non-text input and responded appropriately. Growing rapidly as frontier models go multimodal.
Common in: medical imaging, automotive AV, robotics
Quick answers
Common questions about the process.
How long does approval take?
Some projects allow for near-instant approval. Others take 3–5 business days, depending on the qualifications required. If you don’t pass first time, you can re-apply after 30 days.
Can I work across multiple domains?
Yes. If you qualify in more than one domain, you can access projects across all of them. Many contributors start in one domain and add more over time.
Is there a minimum number of hours?
Each project is unique. Some have no required minimums; others ask for dedicated hours throughout the week.
How do I access emerging vertical projects like robotics or agentic AI?
Apply through the standard process and flag your interest in emerging verticals during your application.
Emerging tech is always evolving — today we might have robotics and agentic AI available to qualified contributors, tomorrow it could be something we’ve never imagined. Once registered, you’ll have access to all of our open job opportunities and can apply as you please.
Ready to start?
The application takes minutes. The assessment is the real filter — and that’s the point.