Index

Building a design system from zero

My role
Product design lead (sole designer)
Team
1 product designer, front-end team
Timeline
Apr – Jul 2025

Description

Turning a freshly launched brand into a working component library that bridges design and delivery, and modernizing the front-end framework in the same move.

Read case study

I built the system first, then evolved it as the discipline changed: phase one was the design system, brand translation and framework migration (April to July 2025). Phase two is the AI-era extension, covered in the design agent.

The problem

I joined just after the company had launched a new brand, but the brand lived in marketing materials, not in the product. There was no component library connecting the new visual identity to what the product actually shipped. Design and delivery had no shared source of truth, so consistency depended on memory and goodwill, and every new screen was a fresh negotiation.

On top of that, the front-end was still built on an older version of its underlying framework. So there were two gaps stacked on each other: a brand that hadn’t reached the product, and a technology base due for modernization.

I saw them as one opportunity. Rather than build a component library on aging foundations, I used the moment to do both at once, establishing the design system and migrating the framework from Material 2 to Material 3, so the new system would sit on a modern base from day one.

What I set out to do

  • Translate the new brand into a visual language that works in a product UI, not just on a landing page. Marketing and product design need to stay aligned in a balanced way, with differences that serve their different purposes.
  • Build a component library that serves as the shared source of truth between design and engineering.
  • Migrate the framework as part of the same effort, rather than as a separate future project.

Process

Translating a brand into a product UI, and holding the line

The first and hardest part wasn’t visual craft. It was a conversation I had to win.

The new brand was built for marketing: gradients, vibrant color, expressive flexibility. That’s right for a campaign and wrong for a working product, where the interface’s job is to get out of the way and let users focus on their content. Stakeholders wanted the product UI to use the exact colors from the brand guidelines. That isn’t possible, because some of those colors can’t carry text or data legibly on top of them.

So I made the case for a disciplined translation instead of a literal copy. I kept the brand’s tone but derived lighter and darker shades from it, working through many variations to guarantee enough contrast and readability in every context the components would appear in. I tested each against established UX and accessibility principles rather than eyeballing it.

The value wasn’t picking colors. It was protecting usability against a well-intentioned brand mandate, with evidence rather than opinion, so the decision held.

Working with engineering

Before building anything, I clarified which framework the team was actually on. It sounds obvious, but it determines everything downstream: a design system built on wrong assumptions about the component base is wasted work. We standardized on the default Angular Material components where possible, with a smaller set of customized components only where the product genuinely needed them. Leaning on the defaults kept the system maintainable and kept design and code close together.

A sit-together collaboration model

As the sole designer, I couldn’t validate the system in isolation. I teamed up directly with the front-end team to test proposals across real pages of the product, so we could confirm the solution held up under every real condition, not only in a tidy design file. That close working style was fast, caught problems early, and was a genuine support structure for a one-designer effort.

Outcome

  • Ship-to-delivery sped up by at least 30% after the framework upgrade and component library landed, according to our front-end lead. I treat this as a directional, engineering-reported figure rather than a precisely measured one, but the direction is clear: a shared component base removed a lot of the per-screen negotiation that used to slow delivery.
  • A consistent, professional product experience aligned with the new brand. The visual identity finally reached the product, translated rather than pasted.
  • Better internal collaboration. The library became the shared reference design and engineering had been missing, so both sides stopped re-deriving the same decisions.
  • A modernized technology base. Moving to Material 3 in the same effort means the product is set up for better performance and longevity, rather than carrying an upgrade as future debt.

Where it’s going now

The system’s next chapter is already underway: extending it into an instruction layer that lets AI tools generate on-system designs. See the design agent.