- Role
- The only designer; led front-end through the MVP build, from Figma to React Native
- Outcome
- Demo-ready MVP, built ahead of the backend on mock data; the startup wound down before the backend launched
- Key calls
- Swipe instead of ratings
- Occasion-based recommendations
- Frosted, native-feeling surfaces
The Problem
Getting dressed is a context problem, not a taste problem. You don’t need more clothes. You need to know what to wear to the 7pm dinner in this weather with that person. WheareAI started from that question.
My Role
Founding Product Designer and Front-End Engineer for WheareAI.
An AI wardrobe and outfit-discovery app built in React Native (Expo). I was the only designer and led front-end through the MVP build, working with the founders and two backend engineers. I took the front end from the first Figma frame to a demo-ready MVP. It was built ahead of the backend, on mock data, and the startup wound down before the backend launched.
Created
2024-2026
Key Design Decisions
Swipe as the primary feedback system
I chose swipe-based FIT / NOT IT over star ratings or explicit tagging because it mirrors how people actually make quick decisions on a phone. Additionally, it generates preference data without asking users to think. The tradeoff was less granular signal per swipe. The plan was to offset that with volume over time, feeding a recommendation model as the user’s closet and swipe history grew together. That backend layer was scoped but never shipped before the startup wound down.
Occasion as the recommendation axis
Rather than a generic “suggest an outfit” prompt, I structured recommendations around context (getting dressed is a situational problem, not a taste problem). The right outfit for a Tuesday commute is wrong for a Saturday dinner. So recommendations start from the occasion: pick one, like Date Night or a Job Interview, and the outfits follow. This shaped the entire information architecture of the Whearables tab.
Camera capture, built ahead of AI segmentation
Adding clothes had to be fast enough that people would actually do it. I implemented the camera capture screen in code (AI-assisted); the plan was for an AI segmentation model to auto-crop and classify each item from the photo. I built the front-end side ahead of the backend (API contracts and an integration layer for the backend team); it wasn’t part of the demo build, and the model never shipped before the startup wound down.
Onboarding Flow
Onboarding had one job: learn enough about you that the first recommendations weren’t random. It runs from phone verification through your name, location and basic info (including body type) to two rounds of FIT / NOT IT swipes: outfits by style, then by occasion.
Whearables & Swipe UX
The whole product hinges on three things: getting clothes into the closet, surfacing Whearables (recommendations), and closing the loop with feedback. I designed and built the main tab architecture (closet, camera capture, Whearables), the interaction patterns (gestures, loading, empty states), and the swipe experience, so recommendations feel fast and legible on a real device, not just in a static mock. To keep the feed fast, I built a WebP image pipeline that cut outfit photos from about 1 MB to 100–250 KB.
Design System
I defined a design system in Figma (type, color, components) and built a matching React Native component library (tokens and shared components), so every screen stayed visually consistent and easy to extend. Design-to-code was in my own hands, so decisions in Figma showed up the same way in the app. The handoff that mattered was to the backend team, so I wrote it down: an API contract, a backend migration plan, and a demo guide for investor meetings.
I deliberately used frosted, glass-style surfaces (blur-backed cards and layered translucency) on key flows, specifically the navigation bar, so the UI felt light and native without maintaining a fully custom chrome. The tab bar follows iOS’s native pattern for familiarity and less engineering overhead.
Social Vision & Roadmap
The MVP was always solo-first, but social discovery was in the roadmap from day one. The profile picture in the wardrobe wasn’t accidental. It was a foundation for future friend-to-friend wardrobes. The plan was a subscription tier where users could browse friends’ closets and save their favorited outfits. Classic social hook: turn a solo utility into a discovery layer, put the highest-engagement feature behind the paywall.
How I worked (and what stuck)
As a founding designer and front-end engineer, the tradeoffs were constant: depth vs. time, polish vs. scope, native feel vs. solo bandwidth. I leaned on clear component patterns, reuse, and AI-assisted coding to build a demo-ready consumer app without a design team, working through tickets and pull requests (my PRs were merged, including one merged by a teammate). If I had another cycle, I’d instrument the swipe feed from day one. I’d want to know at what swipe count recommendations started feeling personal, and design the onboarding length around that number.