Mamo, Design system, 2024–2026
Blaze Design System
Blaze is the design system Mamo’s products are built on. I wrote its first code as an engineer, then moved into design and took over its Figma source. From the design side, my job was to keep it consistent, documented, and in front of the engineers who build from it.
- 140+
- 5% to 100%
- 900K+
The challenge
We had a design system without it really being one. Components carried too many variants, most had no documentation, content guidelines weren’t part of the system, and a change published in Figma didn’t reliably reach the engineers who needed it.
I came at it with an advantage. I wrote and shipped the code before I owned the Figma source, so I know how a system breaks in implementation. The Figma decisions have to hold up in the codebase, not just look right in a file.
How I approached it
Fewer variants, one container
When I took over the Figma source, variants were the biggest drag. Components carried a separate variant for every on and off state, or had boolean properties nobody used, and that bloats a Figma file and slows it down.
- Booleans instead of variants: anything with a toggle can be a boolean property. Icons, helper text and other on and off states became booleans.
- A Card with no style variants, on purpose: a card is one kind of thing, a container for information, and it can look like anything depending on what you’re designing for. A gray card, a white card with a border, a gray card with a border. Restricting a component this generic would limit how freely designers can use the system. The system decides when something needs to be a card. How it looks is up to the designer, as long as the patterns are clearly defined and documented. Padding lives in whatever uses the card, visual properties are instance overrides, and feature-specific layouts belong in local components.
- Tokens in semantic tiers: identifier, status and brand, so a new color has a place to live before anyone asks for it.

Making changes visible
Publishing a change in Figma doesn’t tell anyone. Tag an engineer in a comment and they may never be notified. If they are, they’re busy, and they can’t tell what specifically changed. A Slack reminder helps, but even that gets lost. Dozens of published changes piled up with no reliable signal of what had changed or what needed migrating in code.
So the changelog lives in two places:
- Figma holds the permanent record, because Figma is the source of truth.
- When a change needs engineering work, a ticket goes to the right engineer in Linear. It lands on their board, documented in detail, and the product manager can prioritize it.
That closed the gap between a design change and the code.


Documentation at scale
Our components had almost no inline documentation. Engineers guessed at intent and designers had no reference. I looked at how Uber documents its design system for inspiration. Uber gives every component its own file, which was more than we could take on, so each component gets its own page in the design system file instead. That worked.
Every page follows the same five frames: overview with do’s and don’ts, anatomy and variants, sizing, theme, and usage. They’re written thoroughly enough that a designer and an engineer read the same thing.
- By hand first: I documented the highest-traffic components myself to set the quality bar.
- Then with AI: with a structure this clear, AI has little room to get it wrong. Where it earns its keep is reading the documentation and suggesting what’s missing, like edge cases a component should handle natively, so a designer doesn’t have to detach it to add something.


Coming from the engineering side
I came to Blaze from the engineering side, in my time as a UX engineer.
- 2021, SCSS: Mamo had no component library, and the dashboard was built in HTML and Ruby. I built an SCSS library where the styles were the reusable part, so a component came down to reusing them. It saved the engineering team 7 to 10 hours a week and gave us one source of truth to build from.
- The React move, Storybook and Material UI: when the dashboard moved to React, I rebuilt the library there. Material was an inspiration that we gave our own flavor, and Storybook was the sanity check that components matched Figma.
- What it gives me as a designer: I know where the limits are, because I’ve worked in a very restrictive code base and a very free one.
What came of it
Blaze went from a set of components to a documented design system.
- Cleaner variants, with booleans in place of extra variants.
- A five-frame documentation standard, applied to the highest-traffic components first.
- Content guidelines surfaced for developers too.
- A changelog that puts changes in front of the engineers who need them.
- The dashboard migration: engineers were moving the code base and splitting front end from back end, a very big project. It gave the design system a chance to rethink its components and documentation, which made the migration easier too. The merchant dashboard moved onto Blaze from 5% of merchants in October 2025 to 100% by January 2026.
- Scale: 140+ components.
- Between October 2025 and October 2026, six Figma teams placed Blaze components into files 900K+ times and its variables 1.7M times, and 79% of the component use was in the merchant dashboard. The most-detached component was detached 62 times.
Reflection
What I’m proudest of
- Working through both lenses, engineering and design: I’ve worked in a very limited code base and a very free one, so I know how to work around constraints, and I can bridge the two sides.
- Fixing problems on both sides: I saw problems in the code and fixed them, saw problems in the design and fixed those too.
- Going from a design system in name only to a documented one: components, do’s and don’ts, and content guidelines surfaced where developers work.
- Consistency users can feel: someone moving between two features should see the same design language and the same way of communicating, so two features feel like one company, not two.
What isn’t done
A design system is never finished.
- Motion: I started defining Blaze’s first formal motion language, tested against real product features instead of invented in the abstract.
- Icons: I was evaluating a move off Material Icons to a set that matches where the system is heading.
- AI tooling: the open question is what the system looks like when AI tooling can read and extend it directly.
What I would do differently
- Treat the tokens as the product, for every surface: Blaze ended up serving the dashboard, the app and email. I’d name and structure the token layer from the start so web, Flutter and email can all pull from one source.
- Budget for theming earlier: I knew how to structure dark mode, and in Figma the variables are the easy part. The cost is the audit: every accessibility check passing and every existing screen working together, which takes real time and testing. It wasn’t in our time budget, since our merchants are used to light mode and nobody had asked for dark. With Blaze reaching more surfaces, I’d plan for that audit earlier so the option is ready when it’s needed.
- Build it to be used without us in the room: new surfaces and AI tools started asking for the design system, and it had to be explained to them. I’d write the documentation and the tokens from the start so people and tools outside the team can use Blaze on their own.