About Aptura
Aptura is an applied AI and data company working with frontier labs to train AI systems on real professional work. We do this by building high-quality evaluation and training environments that teach AI agents to produce work that can be used in daily workflows.
Role Description
We are looking for Subject Matter Experts from investment banking, credit, structured finance, restructuring or equity research who build merger models, credit, project finance and research models in Excel. You know where a merger model breaks when the financing changes, and why AI rarely gets it right. You have used AI tools on your own work and been frustrated by the mistakes they make, or you simply want to find out how good they really are. You will turn that experience into realistic Excel tasks that test how well AI does the modelling work of a banking or research analyst.
Responsibilities
Writing realistic Excel tasks drawn from your own deal, credit and research work: extending existing models, fixing broken ones and building from a brief.
Working through each task yourself so there is a clear, correct answer
Writing a yes-or-no criterion precise enough that a grader can mark any attempt against it
Necessary skills
3+ years in investment banking, credit, structured finance, project finance, restructuring or equity research
Deep hands-on Excel: purchase accounting, calendarisation, circular references, goal seek, nested formulas and financial functions
Clear writing. Tasks must be written by you, not by AI tools
Attention to detail, and willingness to learn concepts in AI and Evals.
How it works
Apply with your details and CV. After applying, choose an available 30-minute group walkthrough call.
Attend the walkthrough and create one original task from your own work. Submit the brief, the actual input workbook(s) needed to complete it and a write-up of the expected result. Input-file uploads are compulsory.
Write the brief yourself, without AI drafting, rewriting or polishing. Input workbooks must also be human-built. Use real data that has been sanitised, or public data with the source named.
Use only materials you are authorised to share, with all confidential employer, client and personal information removed, including from hidden sheets, comments, pivot caches and external links.
Submit through your account for automated screening and team review. Your task is run against current frontier AI models. If the models get it wrong, you then write a detailed yes-or-no rubric, and a reviewer signs off on the task with you. Follow any request for revisions in your account.
Completing onboarding does not guarantee a job offer; selection remains subject to review of the task proposal.
