About the Role
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 EngineeringWrite, 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:Advanced proficiency in Python, JavaScript/TypeScript, Java, C/C++, Go, or RustA solid grasp of data structures, algorithms, and software design principlesA CS degree or equivalent professional experienceNice to have:Open-source contributions or competitive programmingDistributed systems, cloud, or DevOps backgroundCode review, technical writing, or mentoring experienceTrack 2 — Finance & AccountingAuthor 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:Hands-on experience as an accountant, financial analyst, controller, or auditorWorking knowledge of revenue recognition, matching principles, and standard close workflowsComfort with financial documents, spreadsheets, and accounting system outputsNice to have:CPA, CA, ACCA, CIMA, or CFAIFRS or US GAAP depthERP familiarity (SAP, Oracle, NetSuite, QuickBooks)Audit, financial controls, or revenue assurance backgroundTrack 3 — Medical & ClinicalChallenge 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: