Design context contract for agent-built UI
Designed and distributed an org-wide design context contract that AI agents consume directly, so agent-generated UI stays on-brand without requiring human review on every output.
Design systems are infrastructure that shapes what's possible.
I build design systems that teams and AI agents can both act on. Token pipelines, documentation formats, contribution models, designed for a world where code writes code.
color/brand/primary
Alias: color/palette/blue-400
--color-bg-action-default
Semantic · light + dark
#7A9EE8
Compiled · WCAG AA ✓ (4.6:1)
Selected projects
Markdown-native design system source
Migrated the design system to a markdown source format that serves two audiences simultaneously: the designer reading it and the AI agent parsing it. Component specs, usage patterns, and prefer/avoid guidance, human-readable and agent-consumable.
Semantic token system enabling dark mode and multi-brand theming
Built a semantic token architecture across Figma variables and code: dark mode, brand theming, density variants. Semantic tokens propagate everywhere it matters.
Design patterns
Defined reusable patterns for recurring UX problems above the component level, each documenting approved solutions, explicitly disallowed ones, and guidance on when to use which. A single source of truth that aligns product, growth, marketing, and brand before anything ships.
Figma-to-code parity across the Zinnia component library
Established and maintained Figma-to-code parity for a 90+ component library. Used Figma MCP to automate parity validation, replacing manual sync checks with a repeatable, agent-driven process. The system stays accurate without scaling the team.
The way I think
I got into design systems because of leverage: the places where one well-considered decision ripples out and changes everything downstream. I'm drawn to the structural fix over the local fix.
A token architecture done right doesn't just enable dark mode. A contribution model done right doesn't just scale a library. The decisions compound so the platform becomes the one doing the work.
I use AI agents as actual tools in my practice, not as a novelty. When agents can consume and act on system guidelines directly, the quality floor works even when no one's in the loop. It's a different way of thinking about what a design system is for, and what a platform team's job is.
Outside work I make music, teach and practice yoga, and ride my bicycles around San Diego County. Making music taught me that constraints are generative: a limited palette, a fixed tempo, a key. Design systems work the same way.
Where I've worked
Zinnia Design System
Responsible for the full token pipeline end-to-end, component API design, Figma-to-code parity, and the documentation architecture built to serve both humans and AI agents on Zinnia, Zapier's AI-first design system. I lead contribution models and quality standards across 25+ designers and 280+ engineers. The system serves both humans and agents, and it scales without the need to grow the team.
Liberty Mutual Design System
Embedded on the Liberty Mutual Design System team as both designer and front-end developer, owning the full component lifecycle from visual design through coded implementation. Focused on API clarity and predictable engineer handoff, contributing to documentation consumed by an S&P 500 distributed org.
Product UI & web
Designer and front-end developer for a social viewing platform. Built the full application UI and marketing site, ran user research from interviews through usability testing, and owned design from concept through production.
Working on something where a solid platform has real stakes?
Climate, civic tech, education, art and music. If the system is the work, not a support function, let's talk.
Send an email