← Work · 01 — Trafilea · Internal AI platform · 2025–26

Darwin, where a CRO team keeps what it learns

Trafilea’s UX and CRO teams kept finding problems on the stores and running A/B tests, with nowhere to keep either. With the UX team I designed Darwin, an internal AI platform where a finding goes straight to the developers and every test result stays easy to find. Bugs on our stores now get fixed the same day.

CompanyTrafilea
Year2025–26
RoleSenior Product Designer, designed and built with AI
TeamUX + CRO, ~20 people
Built withLovable, Cursor, Supabase
Swipe for more →
Darwin home: store audit, test conclusions, market watch and hypotheses, with test opportunities below
01 · Challenge

Designers kept finding broken pages, and had nowhere to send them.

Whoever found a problem had to go and tell a developer in private, and nobody could say how long a fix took.

What past A/B tests had taught was locked inside VWO, far from where designers work.

02 · Solution · done together

We shaped the idea as a UX team, then each designer built a version.

We argued about what the tool should be, built our versions in Lovable and presented them. The product took the strongest parts of each.

Mine started from a designer’s question: what do we already know about this page? Three ideas made it into Darwin.

Swipe for more →
Luiza's first draft of Darwin, built in Lovable
My first draft, built in Lovable
Tests by page

Collection, product page, checkout, homepage, each with its own history.

Why it won

A short explanation in plain words next to each result, because a lift alone teaches nothing.

Gaps

Notes like “two months without a product page test”, to show where nobody is looking.

03 · How it was built

No developers. Designers shipped it with AI.

After the Lovable versions, we moved the product to Cursor, connected it to Supabase and put it live ourselves.

No developers and no other teams involved: the UX team designed, built and shipped Darwin with AI as the tool.

Balance people & business

A problem found on the store goes straight to the developers’ Slack.

It keeps its status, so everyone sees what is open, being fixed or done. A private chat became a workflow every team can follow.

Swipe for more →
Store audit screen: findings with priority, status and a Send to Slack button
Test fast · the part I own

Darwin drafts the next test and shows where the idea came from.

This is the part I own. Every suggestion names the past tests and references behind it, so a designer can check the reasoning before spending a test slot.

Swipe for more →
Test opportunity screen: a suggested test with the past tests and references it is based on
04 · Impact

Same day

A broken store page costs sales for as long as it stays up. Now it gets fixed the same day, sometimes within hours.

Before Darwin nobody could even tell how long a fix took. Today the UX and CRO teams work from it, about ten people each, and new tests can start from what the company already learned.

What it doesn’t do

It doesn’t have great ideas on its own. It points where to look; the designer decides.

What’s next

It doesn’t know our customers yet. I now own activation: other teams opening Darwin, getting it the first time, and coming back. Customer experience is next to join.

Screens redrawn with sample data. The real platform holds company data.