A sustainable wardrobe app that helps users reconnect with what they already own — built with AI, computer vision, and a lot of empathy for the people who want to shop less but don't know where to start.
This project grew out of a UX research mentorship at UCSB, where I was exploring friction in everyday consumer behavior. I kept coming back to fashion — specifically the gap between wanting to be sustainable and not having the tools to actually do it.
Most people don't overconsume because they're careless. They overconsume because they've lost track of what they have, forgotten how to style it, and don't know how to start rebuilding intentionally.
"Sustainable shopping feels rewarding — but knowing what's worth keeping, what fits your style, or how to even start building a wardrobe that lasts? That's where people get stuck."
Users reported feeling overwhelmed by their own closets — too many items with no system, no memory of why they bought things, and no way to see what they actually had.
Existing apps gave generic advice. Users wanted something that understood their specific wardrobe, their style, their body — not a one-size algorithm.
Most wardrobe apps were built around shopping and discovery, not rediscovery. They inadvertently reinforced the overconsumption loop they claimed to solve.
I ran usability testing across three rounds with 12 participants, measuring task success rate, error rate, and perceived helpfulness. The scores tracked real improvements as I iterated on the design.
The most surprising finding: users who identified as not fashion-savvy rated the app highest — 4.48 out of 5. The app's value was clearest for people who felt lost, not for those with existing style knowledge.
The solution centers on helping users rediscover existing garments, not buy new ones. Built in Flutter with a Python backend, the app connects to Gemini's multimodal API for computer vision and conversational AI.
Users upload photos of their garments. Gemini's computer vision automatically categorizes and organizes them — by type, color, occasion, season. The result is a visual wardrobe inventory that feels like a new closet, built from existing clothes.
A prompt-engineered chat interface that helps users build outfits from what they already own. The stylist asks clarifying questions, learns preferences over time, and always recommends garment reuse before any external discovery — reinforcing the sustainable behavior loop.
The hardest design challenge was restraint. Every feature had to earn its place — the simpler the interface, the more users trusted it to solve their actual problem.
The stylist's personality mattered as much as its functionality. Users responded to warmth, humor, and non-judgmental framing — especially around the "not fashion-savvy" identity.
Users who felt the app genuinely cared about helping them — not selling to them — engaged more deeply and reported feeling better about their existing wardrobes. Meaning trumped utility.
Adding a short onboarding questionnaire about style goals dramatically improved recommendation relevance — users felt heard before they'd even uploaded a garment.
Users wanted to bookmark outfit combinations they loved — turning the app from a tool into a personal style archive. A simple addition with outsized engagement impact.
With a 4.48/5 average score, users who came in feeling the least confident left feeling the most empowered. The app's real value was lowering the barrier to self-expression, not elevating style expertise.