Inspiration
I wanted a closet management app that doesn't tell you how to dress in a paternalistic way, and could help the fashion conscious and clueless alike. By taking a subject that is inherently opinionated, its a great way to take advantage of how generative AI models the full distribution of fashion, and allows users to create their own personal recommendation system.
What it does
The user takes photos and google Gemini describes them. These are processed as episodes by Zep, which manages a temporal knowledge graph for each user. Users create a record of what they wear and can even see how well new items fit their style objectives when shopping.
How we built it
GCP / Google AI Studio for image embeddings, Vercel for Web. Convex for object, vector storage, and application data. Clerk for user management. Expo for the mobile ports. All in the same monorepo. By selecting libraries like effect and technology like Convex, I set 5.6 sol down a happy path where it could manage software development end to end. I steered prototypes and allowed it to further refine features before porting to mobile with open ended and underspecified prompts.
Challenges we ran into
Major challenges were identifying the friction points that led to testers not signing up or churning. Some users struggled to understand they didn't need a "good" photo because the AI isn't finicky in a way that the are used to computers being.
Accomplishments that we're proud of
The routing of full outfits versus individual items, as well as detecting when items are already part of a user's closet via image embeddings. This made the UI very simple. Likewise, the self-updating user bio.
What we learned
Product sense is as or maybe even more important than programming or system architecture ability. Once the architecture was defined and logging and product analytics were implemented, development had a tight feedback loop that was user experience focused and lighter on implementation details. 5.6 Sol learned how to maintain the codebase so I could graduate to product design.
What's next for Wardrobe
After releasing for IOS and Android, a chrome extension online shopping companion.
Built With
- arcjet
- axiom
- clerk
- convex
- expo.io
- gcp
- posthog
- typescript
- vercel
Log in or sign up for Devpost to join the conversation.