BoutiqueAI
A multi-tenant operating system for boutique retailers.
Catalogue, orders, karigars, WhatsApp commerce, billing and hisaab in one product. I designed the 52-table schema, built the Express API and the React Native app, and still run it.
Every boutique gets a catalogue and checkout built from the products the owner already maintains — no second inventory to keep in sync.
The day’s orders, dues and deliveries on one screen.
Stitching, karigars, trial dates and balances on the order record.
Confirmations and receipts from the boutique’s own number.
The notebook became a ledger.
Orders, stitching jobs and sales write to one record. Dues are derived from it, reminders go out on WhatsApp, and the day closes on a single screen.
WhatsApp became the counter.
I registered as a Meta tech provider and built embedded signup, so each boutique sends confirmations and PDF invoices from its own number — queued through BullMQ with retry and backoff.
Tenancy is enforced at the query.
All 52 tables are scoped by boutique at the data layer, with field-level masks on customer and payment data. The client never decides who can see what.
The catalogue got a recommendation engine.
CLIP embeddings in pgvector rank designs by visual similarity, then business rules, colour, embroidery and occasion signals re-rank the shortlist against each customer’s fit history.
