Carbon DeFi case study — Lee Seligman

The problem

Patent-level complexity made the product feel expert-only

Carbon’s trading engine was defined by a roughly 150-page mathematical patent. There was no reference product and no familiar UX pattern to reuse.

The existing design was cold and aimed only at very professional traders. As a result, fewer users explored strategies, sessions were shorter, and fewer users completed and executed strategies.

The job: turn patented financial mechanics into flows traders could understand, without hiding price ranges, asset exposure or execution constraints.

Research & insights

Two interviews changed the design

Every week I interviewed two groups: long-time users who knew earlier versions of the product, and newer users. Two insights changed the design.

Insight 01
A trader described the value he got from the simulator.
What changedAn “execute strategy” button inside the simulator.
Insight 02
Another trader talked about the charts and information next to the strategy form.
What changedChanges to the information hierarchy around the form.

Key decisions

Complex only behind the scenes

Users should feel smart and confident, with no question marks. If there are too many question marks, we did something wrong.

Organize by user intent, not by feature catalog

Trade-off: Powerful tools moved one level deeper.
Why I chose it: Users could mistake Carbon for a basic swap interface, or miss which strategy fit their goal. Grouping strategy types into three levels, Essential, Intermediate and Advanced, gives every trader a clear way in, from a simple limit order to recurring strategies.

Close-up of the Carbon DeFi strategy menu: Essential, Intermediate and Advanced levels, with Intermediate open to show Limit and Recurring Limit — Strategy types grouped by level: Essential, Intermediate, Advanced.

Show consequences next to the controls

Trade-off: A denser screen.
Why I chose it: The challenge wasn't adding controls; it was helping traders understand what their settings would do. Linking the chart to the strategy form shows the result before anything goes on-chain.

Close-up of the Carbon DeFi price chart next to the Sell High and Buy Low minimum and maximum price ranges — The price chart sits right beside the ranges it reflects.

From simulation straight to execution

Trade-off: Risk of blurring hypothetical and real results, so we kept them clearly separated.
Why I chose it: Traders weren't always sure how a strategy would behave as prices moved. Testing and then executing in one flow turns the simulator into decision support.

Close-up of the Carbon DeFi simulation form: Buy Low price and budget, ending in a Create Strategy button — In simulation, Create Strategy is one step away.

Public productAll screens are from the live, public product.

Explore the live platform

Carbon DeFi home screen with total trades, total exchange and trading strategies stats, popular pairs, trending strategies, and tabs for Basic, Advanced and Professional tools — Home: platform activity first, then a way into each level.

Carbon DeFi Explorer with a search bar, Pairs, Strategies, Distribution and Activity tabs, and strategy cards showing budgets and buy and sell ranges — Explorer: every strategy's budget and ranges at a glance.

Design system & process

Designed from zero, validated in live prototypes

I was the only designer on Carbon. I designed it from zero: every screen, every dashboard and the design system. Alongside design, I ran the day-to-day work with development as a PM.

AI-native workflow. I worked with Claude Code, Figma Make and some Antigravity, and always delivered live prototypes. I structured the design system so it connects to Claude Code through MCP.

Typography

A characterful face for headlines, a clean one for the interface.

Typefaces

Bw GradualHeadlines
Euclid Circular AInterface

Scale

Display · 48Lorem ipsum dolor
Number · 361,234.56
Heading · 18Lorem ipsum dolor sit amet
Body · 14Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor.
Label · 12Lorem ipsum

Colors

One bright action color, clear buy and sell signals, and a calm teal base.

Brand

Primary#00E390 → #00DAFF

Semantic

Buy#00E5AA → #00B063
Sell#FF5C75 → #FF0064
Warning#F67400

Neutrals

Background#315D63 → #000406
Inset#001518
Surface#3E767D → #315D63
Text secondary#F8FFFF · 60%
Text#F8FFFF

Results

Less friction, longer engagement

~28% less friction in complex trading flows, measured by task completion and strategy execution.

~12% improved long-term engagement, driven by the advanced trading and simulation features.

Both were measured from the feature release to six months after.

Making risk visible, without blocking experts

UX spec
Errors users can actually fix
Generic swap errors didn’t show what users could fix and what the system couldn’t quote. I placed each error next to the input that caused it, separated user mistakes from route limits, and disabled the action while an error blocks it.
Proposal
Overlap as a choice, not an error
When buy and sell ranges overlap, both can execute in the same price interval, and a generic warning interrupted traders who wanted exactly that. I reframed it as an advanced opt-in, “Enable Internal Arbitrage”, with the consequence spelled out and no promise of profit.
Proposal
One anchor for two parallel ranges
Entering two related price ranges by hand was repetitive and easy to misalign. I made the sell range the anchor: traders enter one offset, and the buy range is calculated, read-only and stays parallel.

What I learned

Evidence is what brings everyone to agree

I learned to stand behind my decisions.

And that the way I present evidence, from research to live prototypes, is what brings everyone to agree.