About
I don’t trust assumptions.
I test before I design and measure after I ship.
Dean Yoo · Product Designer · Chicago, IL · Available for work
The work here started the same way every time: a product that launched before anyone checked whether it worked. At one of these companies I ran the first usability test it had ever done.
I’m the kind of designer who will defend a decision if the research backs it, and drop it the moment it doesn’t. I study at SAIC, live in Chicago, and share a desk with three cats.
Experience
Jan 2026 – May 2026
FiPet
Product Design Intern, Lead on Quiz Battle
Led the end-to-end design of the 1v1 Quiz Battle feature on a 13-person cross-functional team, from concept through hi-fi prototype, owning every screen and state. Introduced the company’s first usability testing and ran it in Maze to validate the design before the next build.
Oct 2025 – Dec 2025
Biasly
Product Design Intern
Sole mobile designer on the team. Owned every mobile screen end-to-end. Redesigned the core news feed to surface political bias before users read a single headline.
May 2025 – Jan 2026
Urban Creator
Graphic Designer
Designed responsive digital interfaces using Webflow and HTML/CSS. Refined component-based design patterns and updated design system documentation.
Research
Every project here has a number attached because every project was tested. Four studies, two of them run twice, where the second round existed because the first one failed. I report distributions rather than averages, say what a sample of five can and cannot prove, and grade a source before it is allowed to influence a decision.
Sift
12 CX professionals
Unmoderated, Maze · two rounds
Round 1 buried the only automation control in Settings. 83.5% misclicked hunting for it.
Round 2 surfaced it as four entry points in the inbox. Misclicks went to 0%, felt control from 3.0 to 4.4.
Biasly
12 readers, ages 22 to 45
Moderated · same cohort, both rounds
69% could not name the bias of an article they had just finished reading.
The indicator moved above the headline. Recognition 31% to 78%, time to recognise 9s to 3s.
FiPet
10 interviewed · 22 sessions recorded
Interviews, then unmoderated Maze
Task success was 100%. Misclick rate was 65.9% and the first screen was read as a buffer.
Success rate was the metric hiding the problem. The other three numbers are what shipped the fix.
Lyft
10 riders
Unmoderated, Maze · concept test
Riders were not asking for a better map. All of them described the same failure: arriving at a station that was not what the app had shown.
Surfacing availability on arrival took confidence in finding an open dock to 4.4 of 5, on a flow every session completed.
Skills
Design
Product Design · Interaction Design · Information Architecture · Design Systems · Usability Testing
Tools
Figma · Adobe Creative Suite · Maze · Framer · Webflow · Protopie
Code
HTML/CSS · React
AI
Claude Code · Google Stitch · UX Pilot · AI-assisted design workflows
This site is built and shipped with Claude Code. Next.js, TypeScript, Tailwind, Framer Motion.
Education
Dec 2026
School of the Art Institute of Chicago
BFA, Visual Communication