ReWear

Inspiration

We started with Ecovolt's challenge: sustainability nudges exist everywhere, but people have become desensitized to them — it's an engagement problem, not a data problem. At the same time, we kept coming back to a very specific, very relatable frustration: buying a piece of clothing, only to realize once it arrives that you already own something almost identical. Textile waste is a huge, under-discussed sustainability issue, and it comes down to a simple gap — people don't have visibility into their own closets at the exact moment they're deciding to buy something new. We wanted to build something that closes that gap, but in a way that's genuinely fun to use every day, not just another "eco-guilt" app that gets deleted after a week.

What it does

ReWear is a digital closet that helps you buy less and wear more of what you already own.

-Scan your closet — snap a photo of any clothing item, and AI vision automatically identifies its category, color, pattern, style, and condition, adding it to your digital wardrobe.

Should I Buy This? — before purchasing something new, scan it. ReWear compares it against your existing closet and flags if you already own something similar, offering a secondhand alternative link instead of encouraging a fresh purchase.
Daily check-in — each day, tap a few items you're wearing from your closet grid. It takes under 30 seconds and builds a streak, turning wardrobe mindfulness into a daily habit rather than an occasional afterthought.
XP, streaks & leaderboard — points are earned from daily check-ins and from choosing secondhand or skipping a purchase, feeding into a streak counter and a leaderboard — a Duolingo-style loop that makes sustainable choices feel like a game you're winning, not a chore.
Closet nudges — items that haven't been worn in a while are flagged with a gentle prompt to donate or repurpose them.

How we built it

We split the build across three tracks running in parallel. One teammate built the vision pipeline: a Node.js/Express backend with an endpoint that takes a photo, downscales it for efficiency, and sends it to a GPT-4o vision model with a tightly-constrained prompt that returns structured JSON (category, color, pattern, style tags, fit, material, dominant color hex code, and more), with retry logic for malformed responses. Another teammate built the data layer, a database for closet items and users, a rule-based similarity scoring function to detect wardrobe duplicates, the secondhand-link generator, and the XP/streak/leaderboard logic that powers the daily habit loop. The third teammate built the frontend in React with Vite, wiring together five screens - closet grid, add item, daily check-in, buy check, and leaderboard - into one cohesive, demo-ready experience. We used Git branches (one per person) merging into a shared main, agreeing on our JSON data contract upfront so all three tracks could build independently and connect smoothly at integration points.

Challenges we ran into

Realigning scope was our first real challenge — our original concept spanned five different hard engineering problems (avatar rendering, real-time wear-tracking, damage detection, live marketplace integration, and more), and we had to be honest early on about what was actually achievable in a few hours versus what needed to be simplified, seeded, or cut entirely. We also realized partway through that our first draft lacked any real reason for someone to open the app daily; it was a useful tool, but not a sticky habit. This pushed us to redesign around a streak-and-XP loop as the actual core engagement mechanic, not an afterthought bolted on at the end.

Accomplishments that we're proud of

We're proud that we built a working, end-to-end AI vision pipeline that reliably turns a real photo into structured, usable data — tested against real-world lighting and background variation, not just one lucky demo photo. We're equally proud that we caught our own engagement gap mid-build and were willing to redesign around it rather than ship something we knew wasn't sticky. And we're proud of how cleanly the three of us split and integrated independent workstreams under real time pressure, without stepping on each other's work.

What we learned

We learned that in sustainability products, the hardest problem usually isn't the data or the technology. It is designing something people actually want to open again tomorrow. We also came away with a much sharper sense of how to scope aggressively under a deadline: cutting a feature isn't a failure, it's what makes the features you keep actually work well. On the technical side, we got hands-on practice with vision-model prompt engineering, structured JSON extraction, and coordinating a multi-person Git workflow in real time.

What's next for ReWear

Next, we plan on expanding this into a social platform. Having real multiplayer leaderboards among friends, team-based streak challenges, and shareable "impact" stats would lean further into the social competition mechanics that make habits stick long-term.