You prompt an AI model to generate and refine real design work, then raise its bar by evolving the underlying design system: a coherent token foundation (color ramps, type scale, spacing, radius, stroke rules), tighter component treatment, and a direction the whole asset follows. You evaluate each output on design-system coherence and visual quality and how well it delivers on the brief, and you direct across multiple rounds of AI-assisted iteration.
What you'll do
Prompt an AI model to generate and refine design work from a brief across multiple turns
Evolve the underlying design system: a refined color ramp, type scale, spacing and radius tokens, and component treatment the whole asset follows
Evaluate each output on design-system coherence, visual quality, and fit to the brief, and flag where adherence breaks
Compare outputs, pick the stronger one, and write the rationale and reproduction prompt to support your expert judgement
Vote and rank preferences to help train models on what a cohesive, high-taste design system looks like
Required skills
4+ years building, maintaining, or applying design systems professionally (in-house design-systems team, agency, or design-engineering role)
You have personally built and owned a token-based system (color ramps, type scales, spacing and radius tokens, components with variants and states), not just consumed one
Sharp eye for adherence: you catch when a radius, casing rule, type step, or spacing value drifts from spec
Structured, specific creative direction grounded in design-systems principles, and clear prompts that direct AI tools across multiple turns
Comfortable reasoning about design systems expressed in code (HTML/CSS/tokens) and spotting where a built interface breaks its own system (React or Figma variables a plus)
Comfortable with AI design tools (Claude, Figma AI, and similar)
Fluent English with strong written expertise; available for async, remote contract work, ~15+ hrs/week
Apply for this role
1Application formTell us about yourself and your work.