What the work actually involves

Projects vary, but the shape is consistent: you produce or review a piece of banking work product, then document the judgment behind it in enough detail that a non-banker grader can apply it. That might mean building a three-statement model and an LBO from a filing set and writing the rubric a grader uses to mark a model's attempt wrong. It might mean reviewing model-generated comps and precedent transactions and explaining precisely where the multiple, the calendarization, or the peer set is off. Or grading a model-drafted CIM or pitch page against what an MD would actually send, with written reasoning behind each deduction.

The written explanation is not a footnote to the work — it is the deliverable. A correct answer with no stated reasoning is close to useless to a research team, because they are training on the reasoning, not the number.

What the platform screens for

  • Hands-on recency. Whether you built the models and wrote the materials yourself, recently enough to explain the judgment calls behind them. Firm name matters less than that.
  • Mechanical depth under follow-up. Expect the AI interviewer to push past the headline answer into circularity, calendarization, stub periods, treatment of NOLs, bridge items in equity value.
  • Evaluation judgment. Can you distinguish a genuine error from a defensible alternative convention, and grade accordingly?
  • Tolerance for ambiguity. Task instructions are sometimes incomplete; the platform wants people who flag that rather than silently guess.

Logistics and honest expectations

This is a standing listing, not a job opening. Applying puts you in the investment banking pool; it does not produce an accept or reject decision. Work is fully remote and asynchronous, with project-based hours rather than a schedule. Past rates have run $100–$130 an hour, observed and set per project by scope and depth — not guaranteed, and named on the specific project listing you're invited to. The initial step is a resume, a confirmation of where you work from, and an AI interview of roughly 20 minutes. Matching can take a week or several months depending on what clients are running.