AI Product Designer

We're changing how product gets built here. Instead of every flow going through a design pass, our engineers assemble from a maintained component system, increasingly with AI coding agents doing the assembly. That only works if the system is genuinely good: opinionated, well-documented, hard to misuse. You'd own the design side of that system end to end. Our frontend team owns the implementation, meaning the library's architecture, releases, and engineering quality. You define what goes in it and why. It's the first hire dedicated to this, so you're building the system, not inheriting it.

Responsibilities:

• Audit our existing surfaces and extract the real primitives: what's a component, what's a variant, what should never have existed.
• Define the token layer covering type scale, spacing, colour, elevation, and motion. Make the decisions so nobody downstream has to.
• Specify components to a level frontend can build from without guessing. Every state, every edge case, the composition rules.
• Write the usage guidance: when to reach for what, what not to do. This is read by engineers and by coding agents, so it needs to be precise rather than aspirational.
• Keep Figma and the shipped components in sync, and treat the shipped version as the truth when they disagree.
• Review what ships across our product teams for system conformance, and fix the system when the same violation keeps recurring.
• Work with frontend on the unglamorous half: what a breaking change means for consumers, how a deprecation gets migrated, accessibility.

Requirements:

• 3-5 years in product design, with at least one design system you owned through real adoption. Multiple teams, multiple releases, actual maintenance.
• Strong opinions about constraints. A system that permits everything hasn't decided anything.
• Comfortable in and around code. You can read a component's props and understand what it can and can't do, navigate a repo, and check whether what shipped matches what you specified. You don't need to be the one writing it, but you'd rather look than ask.
• You use Claude Code or similar agentic tools in your work, and have a view on what they get wrong.
• You can spec at a fidelity that removes ambiguity, and you notice when your own spec is underspecified.

Nice to have:

• Something public we can look at: a Storybook, a repo, a shared Figma library, a side project, a system you've written about.
• Enterprise or B2B product experience, especially where accessibility and compliance are procurement requirements.
• Experience with dense, tool-like interfaces rather than marketing surfaces.
• You've shipped frontend code yourself, at any level.