Amazon
2013 — 2020
The Bridge Team
Sr. Design Manager, Team Lead
I founded the team that kept design intact between the mock and the shipped build across every Amazon device — Fire Phone, Fire TV, tablets, Kindle, Alexa.
Evidence
+121%
Production capacity
Amazon Devices · 2015 → 2019
+290%
Weekly ticket throughput
Design production · 2015 → 2019
01 / The handoff problem
You cannot hand an engineer a redline for a rigged 3D asset.
The Fire Phone's interface was geometry. Every element was modeled, textured, rigged and bound to Android, generated in real time at sixty frames per second and shifting with the position of the user's head. A spec sheet of dimensions and hex values describes none of that. The document that had always carried design into engineering stopped working, and it stopped working in the middle of Amazon's most ambitious hardware programme.
So we split the job. Design execution went one way, documentation went another, and a small group took ownership of the second half — keeping what shipped in agreement with what was designed. It was a workaround under launch pressure. It was also the first version of something I would spend the next seven years building.
That group had no name yet. What it had was a method: separate the people producing design from the people specifying it, and give the specification its own standards, its own tooling and its own quality bar.
02 / The audit
Before I asked for a team I went and counted. I reviewed the design-to-engineering pipeline across every device line in the Digital Devices Group — Kindle, Fire Tablets, Fire TV, and the programmes that never shipped. The same four failures appeared everywhere.
Shipped interfaces diverged visually from the approved mocks. Asset libraries were inconsistent, so the same icon got drawn three times by three teams. Redlining was manual, which made it slow and made it wrong. And there was no central documentation and no QA checkpoint, so nobody found out until a customer did.
None of that is unusual. What made it worth acting on was the direction of travel: the group was adding device programmes faster than it was adding process, and every new programme inherited all four problems on day one. The friction was not going to stay proportional. It was going to compound.
The audit is the part of this story I would repeat anywhere. It turned a complaint into a number, and a number is what you can take to a vice president.
03 / The team
I established the Bridge Team as a permanent cross-functional group embedded inside the product development cycle rather than consulted by it. Five disciplines: design production, design documentation, UX writing, system design, and visual QA — a practice I built from nothing, and the one that turned out to matter most.
The rule was simple. If the group designed it, we supported it. Fire Phone, then Fire TV at launch, then Kindle and Fire Tablets, then every Alexa-powered device from the first day one existed, and a long tail of hardware that never reached a customer.
Embedding mattered more than headcount. A service team that receives tickets arrives after the decisions are made and can only report problems. A team sitting inside the cycle catches the same problems while they are still cheap, and it accumulates enough context to say which ones are worth fixing.
Over four years the Bridge Team stopped being a resource the organisation used and became infrastructure it relied on. That is a harder thing to build and a much harder thing to remove.
04 / Automation and scale
Growth was the forcing function. Programmes arrived faster than I could add people, so the only way to hold the line was to stop doing by hand what a tool could do.
We built automated spec tooling to replace manual redlining, and embedded metadata directly in components so a specification could generate itself from the design file rather than be transcribed out of it. Assets moved into Git, version-controlled, with naming conventions enforced at check-in. Component libraries and design tokens replaced per-screen authoring.
Out of that came the design system for the devices group — the organisation's first, and a shift from static guidelines to reusable components with behaviour attached. It grew to cover Fire TV, Fire Tablets and Alexa. Tooling moved from Sketch with custom plugins to InVision, and finally to Figma, each move chosen for how it let roles work together rather than for the drawing.
The volume that produced is the part people find hard to believe. In the second half of 2019 the team closed more than six thousand tickets. When an urgent requirement landed for Arabic and right-to-left support across every Alexa device, we audited every icon and image, redrew the mirrored and culturally specific assets, and shipped with engineering in under three months.
Automation did not make the team faster in the sense of hurrying. It made the expensive parts cheap, which is a different thing, and it lasts.
05 / Visual QA
The last discipline is the one I would put first if I were writing this today.
Visual QA was two people. They built pixel-level image-diff tooling that compared the shipped build against the approved specification, screen by screen, and annotated every divergence. Each defect was logged as a ticket and tracked to resolution — not reported, tracked. Over the years those two people tested thousands of screens across every product in the portfolio.
The work is unglamorous and the reasoning behind it is not. A design system tells you what should be true. It does not tell you whether it is true in the build a customer will install. Without a verification step, a system is a statement of intent that degrades quietly, one release at a time, and nobody can say when it started.
That is the same problem I have worked on ever since, in different materials. Rules encoded into the design tool so the check happens in the canvas instead of in a review. Systems that have to describe how a product behaves when part of it composes its own interface. Building the conditions in which a design decision survives contact with engineering is not operations. It is the work.
Reflection
The Bridge Team was not founded. It accreted, over seven years, out of a workaround for a phone. That is the honest version and it is the more useful one.
What made it work was never the org chart. It was that every piece got built in response to a failure somebody could point at — a wrong pixel in a shipped build, an icon drawn three times, a specification transcribed by hand and transcribed wrong. Nothing in it was designed in advance, and nothing in it was arbitrary.
What I would defend is the sequence: count the problem, name it precisely, build the smallest thing that fixes it, then make that thing standard. Every part of the Bridge Team came out of that loop. It is the only method I have found that survives an organisation growing faster than its process.
Colophon
Set in Newsreader and DM Mono
Page 1120 · rail 180 · measure 640
Built in Framer
September 2026