Paid AI training work
Braintrust AI training jobs and referrals
A software engineer can earn on Braintrust by checking an AI-generated patch for bugs. A bilingual reviewer might compare two answers and explain which reads more naturally. Braintrust pays contributors for project work. Its Connector Program pays separate rewards to people who introduce qualifying clients or talent.
- Best suited to
- Individuals with domain expertise or a professional hiring network
- Published starting point
- Contributor profile and skills interview for AI training work
- Revenue model
- Project pay for contributors. BTRST rewards for qualifying marketplace referrals.
- Main decision
- Will you do the work yourself or introduce someone who will?
Jump to a section
Who it’s for
Who is Braintrust’s data program for?
Braintrust matches specialists with teams developing AI models. Someone with accounting experience might spot a model that counts revenue twice. A developer might design a test case that exposes a faulty answer. The useful skill in both cases is explaining why an answer succeeds or fails.
For businesses commissioning new work, Braintrust describes how it recruits, evaluates and manages contributors on its enterprise training data page. If your goal is to license an existing company archive, see the micro1 company data profile.
Get prepared
What should you prepare?
Choose a task you could evaluate from experience. Build a short sample that lets a reviewer see your reasoning.
- Name your specialty: Describe a task precisely, such as reviewing Python tests or editing Spanish product documentation.
- Create a sample: Use a prompt and two possible answers. Identify a specific error and explain how you would correct it.
- Complete your profile: Braintrust asks applicants to describe their specialization and complete an AI skills interview.
- Read the project brief: Check the unit of pay, expected review time, revision policy and available volume before accepting an assignment.
Use the effective hourly earnings calculator to include assessment and revision time when comparing offers.
Revenue & payments
How does payment work?
AI training work
Braintrust advertises $20–$45 an hour for labeling and $75–$200 for code review and generation. These are ranges for different tasks. Check the rate on the assignment you can actually accept and how much work is available. Braintrust describes weekly payment processing on its AI training jobs page.
Marketplace referrals
The Connector Program advertises rewards in BTRST tokens. For a qualifying company referral, it lists 2% of paid client invoices, up to $100,000 in BTRST per company. For referred talent, it lists 1% of qualifying invoices, with a separate cap per person. The program tracks introductions through a personal referral link. Review the referral terms before making an introduction.
AIR customer introductions
Braintrust publishes separate terms for referring buyers of its AIR interview software. A qualifying warm introduction can earn a one-time $1,000 bonus. The terms specify at least $10,000 in expected or contracted first-year fees, payment of the first invoice and 30 days of active service afterward. Check the AIR referral terms and eligibility restrictions before introducing a buyer.
Example deals and partnerships
The examples below distinguish delivered projects, reported payments and announced partnerships. Read the evidence label, commercial terms and source for each.
Provider-reported client project
Contributor screening for an AI training data provider
Braintrust reports helping an unnamed AI training data provider onboard over 4,000 vetted contributors per month.
Data, value & context
Braintrust reports using its AIR interview software to assess more than 10,000 candidates per month for an unnamed AI training data vendor. It says the client onboarded over 4,000 vetted contributors per month into active projects.
- Data involved
- Candidate assessment and onboarding for human contributors to AI training projects.
- Public value
- The client, contract price and individual pay rates are undisclosed.
The figures describe how many people the vendor screened and onboarded. Braintrust does not disclose a sale of an existing dataset or earnings per contributor in this example.
Checked . Examples are not earnings estimates.
Rights & control
What can you share in an assessment?
Write a fresh work sample using invented or public inputs. A useful code-review sample can be a ten-line function you wrote yourself, a failing test and a short explanation of the fix. A real customer ticket or employer repository can contain material you lack permission to share.
- Who owns the finished work and can you show it in a portfolio?
- Which reference materials and AI tools does the project allow?
- How does the client handle revisions or rejected submissions?
Explore your options
When should you compare another platform?
For another individual route into model evaluation, see the Handshake AI Fellowship. To compare platforms by task, assessment requirements and time spent finding work, use the guide to earning money training AI.
Good questions, clear answers
Questions before you apply
What would I deliver on a training project?
It depends on the brief. You might rank responses, annotate an example, write a question with a known answer or test whether generated code works.
Can I earn a referral reward by sharing the public jobs page?
The Connector Program directs participants to use the personalized link in their account. Check that the introduction qualifies and that Braintrust can credit it before relying on a reward.
Which program applies when a company wants interview software?
Braintrust AIR has its own customer-introduction terms. A company hiring freelancers through the talent marketplace follows the Connector Program rules.
Sources and editorial approach
Provider information checked on 5 November 2026. Practical checklists are Data Mining Masters’ editorial guidance. This profile is based on public information, not a firsthand transaction or a guarantee of acceptance or income. How we compare platforms.
Keep exploring