Inspiration
Fashion advice usually sits behind a stylist’s fee or a lot of trial and error. People own clothes they never wear, buy pieces that don’t suit them, and can’t tell how something will look before it arrives. We wanted a personal stylist in your pocket that knows your wardrobe, your tone and fit, and your taste. It should let you see an outfit on yourself before you buy it, so shopping feels confident and affordable.
What it does
EVOQE is an AI-powered styling and fashion discovery app that brings together:
Digital wardrobe: organize what you own and build outfits from it. Virtual try-on: see clothing, and [makeup/accessories], on your own photo. AI styling assistant: outfit suggestions and answers to styling questions, based on your profile and wardrobe. Aesthetic lookbooks: explore styles like [list your aesthetics] and get looks assembled for you. Smart shopping: a product catalogue pulled from Shopify stores, tagged by aesthetic, so recommendations are relevant and purchase decisions are better informed.
How we built it
Frontend: a Flutter app with an editorial-inspired UI, structured session and state layers, and a local database for accounts and profile data. Backend: a Python service that calls the Shopify, YouCam (Perfect Corp) and LLM APIs and exposes its own REST API to the app. Catalogue pipeline: products are ingested from demo Shopify dev stores into a database. An LLM assigns each one an aesthetic tag before it’s served. Try-on: YouCam virtual try-on APIs, with [LLM-generated structured inputs derived from product details]. AI features: an LLM-powered assistant and lookbook generator using the catalogue and the user’s profile. Privacy: hashed passwords, and encryption for sensitive profile data like body photos and measurements.
Challenges we ran into
Turning product listings into try-on inputs. Descriptions are free text, but try-on APIs need structured parameters like colors, textures and categories. We had to design a conversion step with validation. Making AI tagging consistent. Getting an LLM to classify aesthetics reliably across messy catalogue data took iteration. Handling sensitive data responsibly. Selfies and body measurements are personal, so we built encryption and secure handling in from the start. Working with async and paid API calls. Try-on jobs take time and cost credits, so we designed around polling, caching and rate limits.
Accomplishments that we're proud of
A complete experience in one app, from wardrobe to try-on to shopping, rather than a single-feature demo. A working pipeline from live Shopify catalogues to AI-tagged, try-on-ready products. Privacy-conscious handling of sensitive user data from day one.
What we learned
How to structure LLM outputs reliably (schemas, validation, retries) so other systems can depend on them. How virtual try-on APIs work, and how strongly input quality affects results. How to design a Flutter app with clean state and session architecture, and a backend that keeps API keys and logic off the client. That personalization is mostly a data problem: good tagging and profile data matter as much as the model.
What's next for Evoqe
Real accounts and cloud sync, with a production backend. Expanding the catalogue to more stores and brands. Smarter recommendations that learn from wardrobe usage and feedback. More try-on categories, and shareable lookbooks and outfits. Price tracking and “wear what you own” suggestions to make fashion more affordable and sustainable.
Built With
- aes-256-gcm
- flutter
- gemini
- geminiapi
- postgresql
- python
- shopify
- supabase
- youcam
