Back to Homepage

Enabling AI-generated Semantic UI across 30+ Apps

I designed a framework-agnostic design system that helps engineering teams and AI agents generate consistent, production-ready UI across 30+ healthcare applications.

Summary

I designed a framework-agnostic system that helps product teams and AI agents generate consistent, production-ready UI across 30+ healthcare applications.

Inconsistent component implementations slowed development and fragmented the experience across 30+ healthcare applications. As lead designer, I developed a design-system architecture that gave both product teams and AI coding agents the context needed to produce consistent, maintainable UI.

The system encoded design decisions through shared tokens, composable components, implementation rules, and documented patterns.

Process

Assess Current State → Establish Foundations → Build Components → Enable AI → Scale Adoption

Current State

An internal survey between engineering and product, concluded that 88% of our developers are writing front end code and 65% of our developers are NOT Confident when something had been developed before.

"What we call our design system is actually just a collection of random components with no consistent API or implementation strategy." Lead Engineer

Foundation Building

We created shared tokens and a framework-agnostic theming layer that supported Angular, React, Web Components, and Dart. The architecture also maintained compatibility with the legacy Angular design system, allowing applications to migrate incrementally.

Sequence of Variables and Components

JSON Tokens → CSS Variables → Component Tokens and Themes → Product-Specific Components

Component Development

Granular components, shared foundations, and explicit implementation rules reduced the decisions AI agents had to infer. We also created Copilot modes for planning, implementation, and code review.

Atomic Components

Ions (CSS Variables) → Atoms → Molecules → Organisms → Templates

Data Table Cells Example

String Cell → Hyper Link Cell

A string cell provides the foundation for a reusable hyperlink cell.

Data Table Functions and Paginator Example

Filter Button + Search + Menu Button → Table Functions

Chip List Box + Chip List Box → Paginator

Composable controls create reusable table functions and pagination patterns.

Generic Table Card Example

Documented composition patterns helped AI agents assemble reliable interfaces and extend the system without introducing one-off implementations. Granular components allows for better component tree-shaking.

Button + Icon Button + Slide Toggle + Table Functions + Table Cells + Paginator → Generic Table Card

A configurable table template assembled from shared atoms, molecules, and organisms.

Scaling Adoption Through Figma and Storybook

We supported adoption across 800+ developers through Storybook examples, architecture decision records, demonstrations, and recorded updates.

Results

Within my specific product, eight engineering teams and cloud AI agents adopted the system. Over the following quarter, team velocity increased 65% while established quality standards were maintained (%99.99 platform availability.)

Teams now use the system to generate scalable, production-ready interfaces from requirements and visual references.

‍

Connecting Figma and Code

Generating Semantic UI from Screenshots

The design-to-code pipeline continues to evolve. Read how WellSky is approaching agentic AI in software architecture.