AI trainer and data annotation jobs
Remote, freelance work helping AI models get better in your language: rating answers, writing prompts and model responses, labeling text and audio, and testing chatbots. Native speakers of many languages, and people with subject knowledge, are welcome to apply.
- Work
- Rating, writing, labeling, transcribing and testing
- Where
- Remote, from your own computer
- Who
- Native speakers, and people with subject knowledge
- Terms
- Rates and terms shared per project, after you apply
What AI trainers do
AI models learn from examples, and many of those examples come from people. AI trainers read model answers and score them, write better answers, label text and audio, and try to make chatbots fail so the problems can be fixed. Work in languages other than English is in demand, because most models are weakest there.
Each project comes with written guidelines and a short qualification task. You work on your own schedule within the project's deadlines, and a reviewer checks a sample of your work and gives feedback.
The kinds of tasks
Projects vary. Most use one or two of these.
Rating AI answers
Score answers for accuracy, helpfulness and how natural they sound, and explain your score.
Writing prompts and answers
Write questions people would really ask, and the answer a good assistant should give.
Data annotation
Label names, topics, sentiment and intent in text, or draw boxes on images.
Transcription and audio labeling
Transcribe recordings word for word, mark speakers and tag background sounds.
Speech recording
Read scripts or hold short conversations for speech recognition and voice projects.
Testing and red teaming
Try to make chatbots give wrong or unsafe answers, and write up what you found.
How it works
Rates, terms and project details are shared when we invite you to a project.
Apply
Tell us your languages, subjects and the tasks you'd like to do.
Invitation
When a project matches, we send the details, rates and terms.
Qualification
Read the guidelines and complete a short qualification task.
Skills check QAOnboarding
Sign a confidentiality agreement and get access to the task tool.
Work and feedback
Complete tasks on your schedule, with reviewer feedback on a sample.
What helps your application
You don't need AI experience. These count for more.
- A native language other than English. Or a regional variety, such as Gulf Arabic or Brazilian Portuguese.
- Subject knowledge. Medicine, law, finance, coding, science or teaching are often needed.
- Careful reading and clear writing. Most tasks need short, precise explanations.
- A reliable computer and connection. Tasks run in a web browser.
Questions about AI trainer work
What does an AI trainer do?
An AI trainer reviews and improves what AI models produce. Tasks include scoring answers, writing example prompts and answers, labeling data, and testing chatbots for mistakes. The work teaches models what good output looks like.
Do I need experience in AI or data annotation?
No. Most projects need native language skills, careful reading and clear writing. Some projects also ask for knowledge of a subject such as medicine, law or coding. Every project starts with guidelines and a qualification task.
Is the work remote?
Yes. All AI trainer and data annotation tasks are done remotely in a web browser, on your own schedule within each project's deadlines.
How much does AI trainer work pay?
Rates depend on the project, the language and the skills it needs. We share the rate and terms when we invite you to a project, before you start any paid work.
Which languages are needed?
Many. Demand changes with each project, and languages other than English are often in short supply, including regional varieties. Tell us every language you speak natively.