D2C Fashion Storefront with AI Try-On
A direct-to-consumer fashion store shoppers can try on before they buy.
- Role
- Backend & product engineer, end to end
- Year
- 2024
- Timeline
- ~10 weeks to first launch, then ongoing
A direct-to-consumer fashion brand was selling through a generic template store. Product pages were slow, returns were high because customers couldn't judge fit online, and the team ran catalog, pricing and orders through a patchwork of disconnected tools.
Fashion catalogs are variant-heavy: every product multiplies into sizes, colours and per-SKU stock. The storefront had to stay fast and SEO-friendly for organic traffic while an AI try-on feature and a returns-heavy checkout ran on top of it.
Split the system into a Django REST backend and a server-rendered Next.js storefront.
Instead of A single monolithic template store.
Why: SSR kept catalog pages fast and indexable for organic traffic, while one API served the storefront, the admin and any future mobile client from a single contract.
Made AI virtual try-on and size recommendation opt-in and cached.
Instead of Running the models synchronously on every product view.
Why: Try-on is expensive compute; making it opt-in and cached kept product pages instant for everyone and reserved the heavy work for shoppers who actually wanted it.
Modelled variants, inventory and pricing explicitly, with faceted search.
Instead of Flattening products into one denormalised table.
Why: Explicit variants made stock, pricing and faceted filtering correct under concurrency, which is the difference between overselling and not.
- Django REST API with faceted catalog search & caching
- Server-rendered Next.js storefront
- AI virtual try-on + size recommendation
- Cart → checkout → payment → fulfilment → returns
- Gift-card and flash-sale engines
- RBAC admin for catalog, pricing, coupons & orders
- AWS: EC2, S3 media, RDS, Docker, Nginx, CI/CD, CloudWatch
Hemant Manglani
Ahmedabad, India
Have a product that has to work? Let’s talk.
A 20-minute call, no pitch. We work out whether this is a real problem worth solving, and if I’m not the right person, I’ll tell you.
Every message answered within one business day