What the work actually looks like

You will spend most of your time doing two things. The first is authoring: constructing a realistic prompt ("a strategic acquirer is evaluating a take-private of a mid-cap SaaS business; build the merger consequences analysis and summarize the accretion/dilution outcome under three financing mixes") and then producing the gold-standard response yourself — the model, the assumptions, the write-up, the caveats an associate would flag before it went to an MD. The second is evaluation: reading AI-generated attempts at those same tasks and marking where they break. Breakage in finance is rarely a wrong sum. It is a terminal value that implies a perpetuity growth rate above nominal GDP, a mid-year convention applied inconsistently, an LBO that ignores the cash sweep, a comps set that quietly includes a company with a different fiscal year end, or a deal rationale paragraph that sounds authoritative and says nothing.

Tasks arrive in batches with a rubric. You score against it, write the justification for your score in a few sentences, and flag cases the rubric does not cover — that feedback loop is a real part of the job, not a formality. Project leads run calibration sessions where several reviewers grade the same item and reconcile differences, and your ability to articulate why an assumption is defensible matters more than whether your number matches someone else's.

What the screen is looking for

micro1 runs an AI-led interview before any human sees your file. It probes deal-level specifics: what you built, which transaction types, whether the models were genuinely yours or you were formatting someone else's output. Expect follow-ups that go one layer deeper than your first answer — naming a technique is not the same as defending a choice within it. Coverage breadth across M&A, LevFin, ECM/DCM or a sector group is useful, since task batches move across mandate types. No AI or annotation background is required or expected.

Logistics

  • Fully remote, contractor engagement, invoiced hourly.
  • Largely asynchronous — you pick up batches against deadlines rather than sitting in a live queue, though calibration calls are usually scheduled.
  • Hours are flexible and variable; some contributors run 10–15 hours a week alongside other work, others take heavier loads when a project ramps.
  • Observed pay band for this listing is $45–100/hr, positioned by seniority, transaction depth, and evaluation quality once you are calibrated. Rates are as observed, not guaranteed.