wardro8e is a mobile fashion discovery app that learns each shopper's style from a short quiz and from what they like, save and view, then ranks products from many stores into one personal feed.
The fashion e-commerce space is dominated by generic, search-based marketplaces like Myntra and platforms like Pinterest that offer visual inspiration but lack direct shopping integration. Traditional platforms overwhelm users with excessive choice and fail to understand individual style preferences.
ByteHint architected a comprehensive dual-platform solution combining mobile-first user experience with robust web-based brand management tools, powered by AI-driven recommendations.
Moves each user from cold start to quiz-based to behavioral ranking as signal builds up, with a 15 minute cache and infinite scroll.
CLIP image similarity blended with style, occasion, colour and price signals to rank items that look alike.
Weighted interactions with a 14-day half-life, views deduplicated per 24 hours, and on-device counters synced in batches.
Two steps covering gender, styles, colours and patterns. Answers seed the feed; skipping is recorded so the app falls back to cold start.
Search, a vibes strip, trending items, community collections and a filter sheet for gender, category, price, colour, size and sort.
Create, edit and delete collections, set them public or private, upload a cover, and save other shoppers' collections.
Saving a product uploads its images, writes the row and stores its CLIP vector in the same request. Editing the main image regenerates it.
Brands sign up with an email code, submit KYC documents and receive a generated partnership contract PDF.
Users browse a Pinterest-style masonry layout of fashion products, with AI learning their style preferences from each interaction (saves, likes, views). Users can explore curated collections, follow favorite brands, and seamlessly purchase directly from independent designers. The platform offers optional style quizzes during onboarding to create a \"style DNA\" profile that evolves with each interaction. For brands, a comprehensive verification process includes business registration validation, document uploads (GST, PAN, trade licenses), contact verification via OTP, and digital signatures on partnership agreements.
System architecture · Architecture
Data model · Data Model
Home feed request with fallback · Sequence
First launch to learned feed · User Flow
Product ingest and embedding · Sequence
Sign-in flow · Auth Flow
Chose A FastAPI service running CLIP through sentence-transformers, called through one HTTP helper with abort timeouts over Ranking in Postgres functions or only in the app
CLIP needs PyTorch, which does not run on the phone or in Postgres. TypeScript fallback scorers mirror the server weights, so a cold or sleeping service never leaves the shopper with an empty screen.
Chose Supabase for Postgres, auth and storage, used straight from the client with the session kept on the device over A custom REST or GraphQL backend in front of the database
A small team and a read-heavy catalog. Auth, storage and Postgres come in one product, and server-only work (embedding writes, KYC, analytics) sits in Next.js API routes.
Chose Local style, colour, pattern, occasion and season counters, with quiz answers seeded at a higher weight, synced on high-intent actions, every 10th view and on app background over Writing every event to the server and recomputing there
Fewer writes, it keeps working between syncs, and the feed reacts right after a like because the feed cache is cleared locally.
Chose A database filter, a service-side filter and a final client filter applied to every source, including cache
A wrong-gender item in a personal feed was treated as a trust problem, so it is guarded at every layer, and an unknown gender for a signed-in user defaults to unisex only.
Chose Bounded noise, a shuffle of the top tier, and caps per brand and per category over Pure score ordering
On a small catalog, strict ordering shows the same items every session. The caps and shuffle keep the feed fresh without dropping relevance.
Phase 1
Phase 2
Phase 3
Phase 4
Phase 5
The wardro8e Android app was built and distributed as an internal APK through Expo EAS, and beta invites were sent to the waitlist. The catalog, admin, brand portal and waitlist site were delivered alongside it.
Tell us what you're building and we'll show you how we'd approach it.
Consumer apps and creator economyA studio promised its creator client an AI social app in a month with one developer. We scoped the real need and had four roles working eight days after the first call.
Gaming and PC optimizationA full-stack, design-first e-commerce build for a PC optimization brand, with cart, checkout and key delivery, that reached a client-reported 25% CTR in a low-conversion niche.
Gaming and PC optimizationA Discord bot that pulls the best gaming deals from Amazon through its Product Advertising API and posts them to the community every four hours.