What the work involves
You are handed a Figma file that an AI design agent has taken partway — some screens laid out, components half-wired, spacing and hierarchy inconsistent in the places where the model lost the thread. Your job is to finish it the way a competent product designer would: repair the component structure, make auto-layout and constraints behave, resolve the states and edge cases the agent skipped, and bring the file to something a developer could actually build from.
Alongside the design output, you are producing evaluation signal. Where the agent's work was wrong, the useful artefact is not just your fix but a clear account of what went wrong and why a designer wouldn't have done it that way — a missing empty state, a text style applied as a one-off override, a component detached instead of variant-swapped. Expect to write short structured notes with most tasks.
What the platform screens for
- Verifiable Figma experience inside a company or agency — not self-taught portfolio work alone. Mercor's screen asks about team file conventions, handoff, and how you worked with engineers.
- UX fundamentals under follow-up: information hierarchy, state coverage, accessibility basics, why a pattern is right rather than merely current.
- Craft specifics in Figma: components and variants, auto-layout, styles and variables, library structure.
- Willingness to critique AI output precisely instead of rewriting everything from scratch.
Logistics
Remote and asynchronous, with tasks claimed from a queue. Observed rates for this project sit in the $75–85/hr band, paid for time logged on assigned tasks; rates are set per project and are not guaranteed. Most contributors work in blocks of several hours rather than short bursts, since finishing a file cleanly is hard to do in fifteen-minute slices. Prior exposure to AI design tools is welcome but not required — the brief explicitly treats learning them as part of the paid work.