How AI-powered visual search and a zero-to-one design system transformed manufacturer discovery for fashion entrepreneurs.
Fashion entrepreneurs struggle to find the right manufacturer. Reddit threads, blog posts, and first-hand interviews uncovered the same friction at every turn — an industry problem hiding in plain sight.
As the founding product designer, I owned the entire design surface — from early research through launch-ready systems.
My background in CS meant I could speak fluently with the engineering team — not just hand off specs and hope for the best.
Make manufacturer discovery feel as intuitive as shopping online — with smart filters, visual search, and progressive disclosure of complex information.
Surface verified information about capabilities, certifications, and past work so founders can make confident choices.
We discovered that the core failure of existing platforms wasn't discovery — it was comprehension. Users knew what they wanted to make; they didn't know how to ask for it.
Fig. 2 · User research synthesis — affinity mapping from 12+ interviews.
Fig. 3 · Information architecture — guided search-to-match journey.
Dual-input search experience — visual hierarchy that guides without overwhelming, progressive disclosure, mobile-first, WCAG 2.1 AA throughout.
Fig. 4–5 · Search interface (left) and Results page (right).
Pomu's engine uses combined image + keyword input to match entrepreneurs with manufacturers. My challenge: make AI feel transparent, not magical.
Fig. 6 · POMU prototype demo — onboarding to results Page.
.png)
I built Pomu's design system from the ground up in Figma — a shared language that made the team faster without sacrificing coherence.
Fig. 7 · Design system — components, typography, grid, color palette.
Since the platform was brand new, I shaped Pomu's identity from scratch. The goal: a visual language that felt bold, modern, and trustworthy.
Every brand decision was tied to a positioning decision — the name, the mark, the palette each had to carry weight with two very different audiences.
Fig. 8 · Brand identity system — logomark, wordmark, and visual guidelines.
We designed LinkedIn templates to ensure brand cohesion across platforms. Result: engagement grew from 0 to 350+ followers after launch.
Fig. 9–10 · LinkedIn campaign templates — primary and secondary content layouts.
Designing Pomu from scratch taught me the power of tight collaboration in zero-to-one environments. I learned to balance brand and product strategy simultaneously while shipping design systems that scale.
Working closely with engineers on the AI pipeline taught me that the most important skill in design isn't taste — it's translation.
Though I've since moved on, the platform's early traction and clarity-first interface remain among my proudest contributions.