Fits App
A fashion-centric social network designed to celebrate personal style and drive fashion discovery — a place to showcase outfits, gain insight, and take inspiration from a global community.
- Role
- CTO & Co-founder
- Duration
- May 2023 – Present
- Users
- 10K+
Stack
Design
Figma for wireframes, mockups, and interactive prototypes across mobile and web platforms.
Frontend
iOS app built with Swift and UIKit, implementing a programmatic UI using SnapKit. Custom components including a waterfall layout for the home feed.
Backend
Scalable backend in Go, with PostgreSQL for storage and Redis for caching. Custom authentication, push notifications, load balancers, an admin panel, and full unit test coverage. Containerised with Docker across staging and production.
DevOps
GitHub Actions for CI/CD into staging and production, Docker images on Amazon ECR, and automated testing workflows to hold code quality across deployments.
Recommendations
Machine learning–driven feed
A model vectorises each image, reading background, colour, text and composition. A custom Postgres image queries those vectors directly, so the feed can surface posts with a similar vibe — connecting people with content that actually matches their style rather than whatever is newest.
Web
A lightweight companion
Recent top fits on the home page for immediate access to what is trending, and profiles that work as portfolios outside the app. Dynamic embeds turn a shared profile or outfit into a rich preview, which is where most of the cross-platform engagement came from.
Impact
Fits passed 10,000 users within a few months of launch, with a global spread that gave the feed genuinely different cultural reference points to draw from. The response centred on two things: the clarity of the interface and the recommendations actually matching people's taste.
The web version and its dynamic embeds moved the needle hardest on reach — a shared outfit stopped being a dead link and started being a preview worth clicking.
Reflection
Building Fits meant owning the whole surface — design, iOS, a Go backend, the ML pipeline and the deployment story — and the most useful lesson was how much of the product lived in the parts nobody sees. The feed only feels effortless because of the vector work behind it.