369 K-beauty products, one inventory number, and an AI skin scan that only recommends what is in stock.

Skincare is a category where customers do not know what to buy. 369 products across 30-plus Korean brands is unhelpful to someone who only knows their skin feels dry or breaks out — in-store staff can advise, but a product grid cannot.

The brand also sells through two channels. Stock sold at a store counter has to disappear from the online storefront immediately, and an online order has to be visible to the team fulfilling it — without two systems drifting apart.

And skincare in Mexico sells through short-form video. The proof that a product works is a creator using it, which is not something a static product page carries.

A storefront on a Django + PostgreSQL commerce backend with Stripe payments, running on AWS, with Celery handling anything that cannot block a request.

Faceted search over the full catalogue — by ingredient, brand, makeup category and skin concern — with live stock state surfaced on every card, down to "only 4 left", because that is what moves a decision.

A shoppable video feed: creator videos with the product attached and a direct path to buy, built into the storefront rather than linking out to social.

A mobile experience that behaves like an app — persistent bottom navigation across Inicio, Productos, Videos, Carrito and Cuenta — because the majority of this traffic is on a phone.

An AI skin analysis pipeline mapping assessed concerns onto the catalogue's existing concern taxonomy, so recommendations regenerate as stock changes instead of being hardcoded.