Portfolio · 2026
Earl
Carlson
Product design leader
About me
I lead design teams on products people use every day.
I give senior designers room to lead, I stay close to the product, and I push the work until it's excellent. Consumer products, connected hardware, platforms, trust and safety, and emerging technology.
Most recently
Senior Product Design Manager, Docker · 2024–2026
Team
12 people at Docker: designers and design engineers
2022–24
A creative sabbatical: music, art, and travel
Based
Ann Arbor, MI · remote
Also
DJ, music, ceramics, surfing

OPTION 3B

My 2010 thesis asked how AI could reshape creative communities. At Docker, I led a design team through it.
Where I've worked
Neighborland
2010–13
Wikia
2013–14
Nokia
2014–15
DigitalOcean
2015–17 · 2019–22
Facebook
2017–18
Docker
2024–26

01 · Docker design team, 2024–2026

Bringing a whole design team into agent tools in a year the rules changed every week

Docker design · snapshot
Role
Senior Product Design Manager
Team
12 people: Staff-heavy designers and 2 design engineers
Surface
Docker Hub, Docker Desktop, agentic products, accounts & billing, the design system
Context
Engineering moved to Claude Code. The rules for design were changing every week.
Real artifact here:
an agentic product screen or
the agent-ready design system
How the team moved, 2025–2026
Earl is designing this one. His account (2026-10-07):
Summer 2025: engineers on Claude Code, still too technical for design → the excited few explore (hackathons; Figma Make, Claude Code, Codex) → clear the old design system's cruft, then rebuild it for agents → weekly show-and-tells → peers pair with slower starters → the new VP makes it the expectation → the whole team.
What I brought
I let the excited designers lead.
Not a mandate, at first: the tools were too early for design. The three or four already experimenting each took a different area, then showed the rest.
I made the learning weekly.
Show-and-tells where someone always showed what they were trying.
I gave the slower starters a peer, not pressure.
Gentler on-ramps, a peer as a coach, and a weekly check-in.
I held the bar on taste.
In 1:1s: "Is your taste in this, or did you just let the agent do it?" [critique practice: what you set up] And I chose not to police every engineering ticket, so I didn't burn out my designers.
The experiments
Screenshot from the agentic
product work
The agentic product surface
One designer, on the product itself
The design system,
in code
A design system built for agents
One designer moved to the system and proposed building it with AI, to be read by agents
Fast trials
Others tried Figma Make, Claude, Claude Code, and v0 to see what held up
The room
Every week, someone showed what they were trying.
Ideas felt real much faster, so we could test them much sooner.
When I knew it had worked
One of our more hesitant senior designers told me Figma Make was slow, and Claude was better.
Once she had an opinion about which agent she liked to use, it had locked in.
What shipped · the design system
Real screens: the agent-ready design system
(a component in code, the docs an agent reads, a screen built from it)
What shipped · agentic products
Delegation and approvals
Sandboxes, staying in control
What changed
~0 → all
designers working primarily in agent tools, in about a year
2
design engineers joined; an agent-ready design system shipped
76%
fewer theme overrides after rebuilding the design-system foundation
"It's not about catching up. It's about shifting the way design works so that the impact of design remains true."
What I was responsible for
Docker Hub
Modernizing it for a world of agents
Docker Desktop
[one line]
Agentic products
Delegation, approvals, sandboxes, staying in control
Accounts & billing
$13.6M net-new ARR · 75K+ migrated subscriptions
Design system
Rebuilt to be read by agents
The team
12 people: 10 product designers (8 Staff) and 2 design-system engineers [how it was organized]
Accounts & billing
$13.6M
net-new ARR from revenue and retention work
75K+
subscriptions migrated
$3.5M
in enterprise churn risk reduced
Sign-in, accounts, teams, checkout, and billing. [one decision design made here]

02 · Docker Hub, 2025–2026

Modernizing Docker Hub for a world of agents and new tools

Docker Hub · snapshot
Role
Design manager and reviewer for Hub Vision
Team
Hub Vision led by Sean Brady, Staff designer
Surface
Docker Hub: search, image pages, and how the Docker ecosystem fits together
Context
A product that had served its purpose, with no major redesign in more than five years
What I brought
Research before redesign.
I set Hub Vision up to understand how people actually navigate Hub before anyone redrew a screen.
Competent wasn't the bar.
The first version was a re-skin. It looked a little better and still held all the content.
I asked what could be removed.
Hub was serving too many masters. I pushed the work from showing everything to showing what matters.
The question I asked
"If you only had three pieces of information to communicate to a user, what would they be?"
What we learned first
Most people never see the homepage. They land on an image page, straight from search.
The old homepage, a catalog of everything, was built for a front door almost nobody used.
93K
sidebar clicks a month [vs. homepage visits?]
57K
same-session continuations
How the work moved
1
A re-skin
It looked a little better. Nothing actually changed.
2
Restructure the data
Closer, and still not far enough.
3
Make the page make sense on landing
The data and the inputs, rebuilt for someone arriving cold.
4
Start from where people come from
Mostly Google, or an agent, not hub.docker.com.
5
Design for agents too
Docker's AI agent in search, and how an agent would use Hub.
I pushed for each of these. The work found its point of view around the fifth, once it started from how people actually arrive.
Before
The image page as it was
After
Page designs
Homepage
Hub for agents
What changed for the designer
Sean started as the person who knew Hub inside and out. By the end, he was quicker to say: it doesn't have to be this way.
My job was to coach, shift priorities, and get it in front of leadership. The pixels were his.
Where it is now
The team is shipping it now, one workstream at a time.
In progress
Homepage searchImage details pageAgent-friendly HubVisual refresh

03 · Facebook, 2017–2018

Fighting misinformation on a feed of a billion people

News Credibility · snapshot
Role
Lead Product Designer
Team
2 product designers · 4 engineers · 1 researcher · 1 content strategist · 1 PM
Surface
News Feed · 1B+ monthly users · web, Android, iOS
Context
A new team, formed the week I joined, as "fake news" became a global crisis
What I brought
I designed the fact-checking tool.
Partners like Snopes rated stories, and a rating reduced a story's reach across Facebook.
I brought publishers in.
I met the New York Times, CNN, and others to introduce the fact-checking and A/B testing tools.
I pushed for credibility signals.
Publisher logos and credibility indicators that landed across the product.
I took the research global.
I condensed field research into product direction, and pushed that our fix was too US- and Euro-centric.
Publisher variants: bringing the source into the story
Credibility indicators in publisher tools: ethics, corrections, ownership
Research · India
"Snopes isn't covering stories in Brazil or India."
Research in Hong Kong, Brazil, and India, where I joined the team, showed the problem looked completely different outside the US and Europe. People got their news in group posts and WhatsApp groups.
Journalist safety
A journalist told us Facebook wasn't safe for journalists to post on.
Cat Audi and I pushed for it, and the team started building safety tools: who can tag you, who sees your posts, what reaches your timeline.
Results
~80%
drop in views once an article was rated as containing misinformation
25
fact-checking partners across 14 countries
Shipped
publisher logos, About This Publisher, and Trust Project credibility indicators
What my designers say
[Testimonial: request from a designer]
Name · Staff Product Designer, Docker
[Testimonial: request from a designer]
Name · Designer, DigitalOcean
[Testimonial: request from a peer or partner]
Name · Role
Let's talk.
Earl Carlson · theearlcarlson@gmail.com · Ann Arbor, MI (remote)