The work

Each task starts with several AI-generated attempts at the same deliverable: a client-facing slide deck, a memo, a financial model, a formatted PDF. You compare them, score each across layout & spacing, typography, colour, domain realism and overall polish, and write justifications that point at specific evidence — this card is 4px shy of the grid, these two columns have unequal heights, the margin discipline breaks on slide 7, the density on this table makes the takeaway unreadable. Then you take the strongest candidate and rework it until it is something a consulting firm or bank would actually send out. The rework is the part that carries the most weight: the model is learning from your finished artefact, not just your score.

What the screen is measuring

The calibration assessment asks you to rate real deliverables and compares your ratings to a calibrated answer key on the layout & spacing axis. It is not testing taste in the abstract; it is testing whether your eye agrees with a defined standard and whether your reasoning is legible to someone auditing it later. Expect the interview to probe for verifiable production history — what you designed, for whom, in what tool, under whose brand guidelines — and for your ability to convert a vague reaction like "this looks off" into a named defect. Consistency matters more than severity: a reviewer who is reliably one notch harsh is usable, a reviewer who scores the same defect differently on Monday and Thursday is not.

Who fits

  • Corporate and in-house designers, agency and marcom people, brand and presentation specialists
  • Management consultants and investment bankers who built the decks and models themselves, not just the content
  • Knowledge-work specialists who are fluent in the deliverable conventions of a specific industry

Logistics

Fully remote and asynchronous, with roughly 15 hours per week of available volume. Work is delivered through Mercor's platform in task batches; there are no fixed shifts, though batches can have completion windows and volume fluctuates between projects. Pay is stated by the platform as $80/hour and is observed rather than guaranteed — the wider $65–100/hr band reflects variation across similar Mercor design projects and role levels.