Creating one shared component library, so four enterprise products stop rebuilding the same UI.

I co-led DesignGrid, the design system behind The Mobility House’s EV charging and energy management software.
- Up to 80%
- less time on component work (estimate)
- 4 products, 1 library
- Dashboard, Installer, Eyond, website
- 2.55:1 → 18.92:1
- contrast on primary table content, clears AA and AAA
- Role
- UX/UI Designer (Co-owner) End-to-end design, system research, user research, design QA on dev builds, user testing, prototyping
- Team
- 2 designers, 4 frontend developers, 1 PO
- Timeline
- Mar 23 – Nov 23
Context
The Mobility House builds the intelligence layer behind EV charging: how energy flows, how loads balance, how fleets charge. Every screen was built from scratch. When marketing needed a website, an agency was brought in. But feedback loops were slow, iterations kept piling up, and nothing was reusable across products. So we took over. They left behind some basics: buttons, drop downs, switches.
Problem
- Components weren’t working for users
- Hotjar showed users struggling with the table. A toggle meant different things in different products.
- Hours explaining, minutes building
- Agency feedback rounds were spent on product context, not design.
- Rebuild, repeat
- Every feature started by tracing patterns from screenshots of old features.
- Expert-only settings
- Brand guidelines applied unevenly. White-labeling meant reworking layouts every time.
- Desktop-first, mobile-broken
- Components that worked on desktop broke on smaller screens.

Getting engineering to “yes”
Built with engineering
Engineering didn’t want to build a library from scratch. Ant Design was how I got to yes. They already knew it, so we shaped it into DesignGrid instead of starting from zero. We structured it using Atomic Design: atoms, molecules, organisms.
Components were designed around dev and user research edge cases. If they wouldn’t reach for it, it wouldn’t ship.

Architecture
1. Atoms

2. Molecules

3. Organisms
3.1 Tables
The problem
a. Hotjar showed users struggling: clicking non-clickable elements, scanning past the data they needed.

b. Tables looked different across every feature. None of them worked on mobile.

What I tried
Letting users pick which columns to show since we had many columns on the enterprise data tables. On phone it was the right call for driver use cases: two columns, user chooses what matters.

Design decision after user testing
On desktop, testers (enterprise admins) didn’t want to configure their view just to see data. We pulled it back to horizontal scroll. The column picker stayed on mobile only.
3.2 Site setup
The problem
1. During testing support team and field electricians flagged site setup as painful. It was buried under “Extended Settings” in the Dashboard, hidden inside a modal. A flat list of fuses and chargers with no way to tell what was connected to what. Users had to click into each component to see specs.
2. None of it worked on mobile, and the Installer app needed it to.

Design decision after user testing
1. Pulled site setup out of the modal and gave it its own tab.
2. Replaced the flat fuse-and-charger list with a collapsible tree: main distribution, sub-distribution, chargers, so the hierarchy matched how the electrical setup actually worked.
3. Color-coded (Chips) consumers and producers (solar and external buildings) so it read at a glance.
4. Added switchable views so specs (Max. Charging Current, Phase Rotation, OCPP, Fallback) showed inline per charger without opening anything.

4. Templates

Testing and validation
Chromatic became the QA layer. Every component change was tested across viewport sizes, interaction states, and edge cases before it reached code.
When we rebuilt the Dashboard with the new components, teams surfaced gaps: “we need a variant for this”, “this doesn’t support that state.” Those requests went back into the library, got tested in Chromatic, and shipped. Build, test, feedback, repeat.
What I learned
Adoption over perfection
I got the system to be adopted by designing around real edge cases users and engineers brought up, not handing off specs and hoping they’d follow. A design system nobody uses is decoration.
Research speaks louder
I spent weeks making the case for rebuilding the tables and got nowhere. Testing with real users and twenty minutes of research on Hotjar session recordings did it. When a decision is stuck, the fastest way out is evidence.
Infrastructure is design work
I arrived thinking I’d design components. What I actually did was reshape how the team makes decisions and collaborate.




