Healthcare UI components don't get to be approximately right. Three and a half years closing the gap between the standard and what the engineer built in good faith.
Terra was Cerner’s open-source design system — the React component library that provided the UI foundation for the Millennium platform. I joined the Terra team three months after I started at Cerner, and I was one of two UX designers on it who could work directly in the React codebase alongside the engineers. I stayed in that role for roughly three and a half years.
Healthcare UI components don’t get to be approximately right. A focus-trap bug in an alert component propagates across every product that consumes it. An icon that fails contrast requirements shows up in operating rooms and emergency departments.
The architecture’s promise — accessible, internationalized, responsive components built once, consumed everywhere — only works if the components faithfully implement those standards. Catching the gaps took someone who could read the design standards and the pull request side by side, and push back on the difference.
Most days I was in the Terra monorepo reading pull requests. Walking through the interaction logic, checking focus management against the standards we’d published, flagging the places where an engineer’s reasonable implementation had quietly diverged from the intended behavior. UX review meant commenting on the PR, citing the specific lines if errors were there, and naming what the standard required. Engineering made the change.
Alongside that, I owned the icon library — a dedicated Git repository the engineering team consumed into the component system, maintained end-to-end from design asset to published package (UX write-up).
The standards we were enforcing weren’t small. The system was built on 125+ design standards and 150+ accessibility guidelines targeting WCAG 2.1 AA. My job was making sure what shipped actually honored them.
The gap between what the standard says and what the engineer built in good faith was rarely about bad standards or bad engineering. By the time I got to Terra, the standards were mature and thoroughly written. The gap was unstated context — the situation in front of the engineer that the standard couldn’t anticipate.
The other thing I learned was personal. I came up on the web believing Jeffrey Zeldman’s line that real designers code — it had been true everywhere I’d worked. Enterprise was where I learned it wasn’t the default anymore. Design and engineering were separate disciplines, and plenty of designers on my team didn’t write CSS, let alone read React. Being one of two designers on Terra who could work in the codebase surprised me — I’d figured that at a tech company code came with the territory, and instead it had become a specialty.
In mid-2022 I moved from the hands-in-code IC role into design management at Cerner, eventually leading up to 12 designers across the US and India. A few months later, Oracle completed its $28.3 billion acquisition of Cerner.
The management philosophy shifted fast: design managers were now expected to deliver meaningful IC work on top of leading. The IC project I took on was the Redwood theme for Terra — a reskin of the component library in Oracle’s Redwood design language so Oracle-built apps could drop into Millennium. Two months in, Oracle Health decided to rebuild its applications from the ground up instead. The bridge became moot.
That ended my Terra chapter. The leadership story and the Redwood design work that came next are each their own case studies. Both start here.
New writing, periodically. No spam, no schedule — just work and thinking as it comes together.