Headquarters: Atlanta, AZ.
URL: https://trainbrain.space/
What if the work you already do every day — shipping code, closing books, reading charts — became the training signal that teaches frontier AI how to think in your field?
TrainBrain builds expert-authored training data and evaluations for AI labs. We're hiring practising professionals to write, solve, and grade the tasks advanced models are measured against. When a model returns a broken function, a wrong reconciliation, or an unsafe clinical suggestion, it's because no expert caught it during training. You'd be that expert.
This isn't your day job in a new wrapper. There are no clients, no month-end close, no tickets, no on-call. You work in writing, on your own schedule, on problems chosen to sit right at the edge of what current models can handle.
No AI background required — we train you on the tooling.
Track 1 — Software Engineering
Write, debug, and optimise code across languages and problem domains. Review AI-generated solutions for correctness, security, and performance, and design edge cases that expose model weaknesses. Provide structured feedback explaining why code works — or doesn't — and compare competing solutions on engineering quality.
You'll need:
Nice to have:
Track 2 — Finance & Accounting
Author reconciliation scenarios: payment-to-invoice matching, billing versus recognised revenue, margin and variance analysis. Build the underlying data and source documents so each scenario is realistic, write rubrics specifying exact figures and required reasoning steps, then solve every task yourself to confirm the numbers hold.
You'll need:
Nice to have:
Track 3 — Medical & Clinical
Challenge models on differential diagnosis, drug interactions, treatment protocols, pathophysiology, and evidence appraisal. Write clinical vignettes with defensible, sourced answers, verify model outputs against current evidence, and document precisely where reasoning breaks down.
You'll need:
Nice to have:
What the work demands
Whatever your track, four things matter more than anything else.
Show your work. Making your reasoning explicit matters as much as the answer itself.
Be verifiable. Every task you write needs a correct answer someone else can confirm.
Iterate. You'll refine tasks until difficulty and gradability are both right.
Flag uncertainty. Saying "I'm not sure, and here's why" is a feature, not a failure.
You'll also need excellent written English, sharp attention to detail, and a secure computer with reliable internet.
Terms
This is a 1099 independent contractor position, not W-2 employment. You're responsible for your own taxes, and company-sponsored benefits such as health insurance, PTO, and retirement contributions don't apply. Contractors outside the US are engaged under the equivalent local arrangement.
Pay: $45–$100/hr, set by domain, seniority, and assessment performance. Specialised and hard-to-source expertise sits at the top of the band, and you're paid on a regular cadence.
Schedule: Choose your own projects, hours, and volume. Scale up in quiet weeks, scale down when your day job gets busy.
Growth: Strong contributors are invited into higher-rate specialist projects and review roles, with ongoing work as new engagements launch.
To apply: https://weworkremotely.com/remote-jobs/trainbrainspace-expert-ai-trainers-in-coding-finance-accounting-and-medicine-nursing
TrainBrainSpace builds expert-authored training data and evaluations for AI labs, enabling frontier AI models to learn from real-world expertise across multiple domains. The company hires practicing professionals—software engineers, accountants, financial analysts, and domain experts—to write, solve, and grade tasks that advanced AI models are measured against. Rather than requiring traditional full-time employment, TrainBrainSpace offers flexible, asynchronous work where experts contribute on their own schedule, focusing on problems at the edge of current model capabilities. The company operates across multiple tracks including software engineering and finance & accounting, providing structured training that helps ensure AI systems can handle complex, real-world scenarios correctly and safely.
