Product architecture
How Lumeira works under the hood
A care-first commerce platform combining a Canadian fulfilment network, an AI try-on engine, and a specialist consultation studio — orchestrated by a single rules engine.
System overview
Four interacting layers — client surface, business logic, data and partner integrations.
Web app, customer/consultant/supplier/admin workspaces
Pricing, eligibility, fulfilment routing
Catalogue, orders, sessions, customers, content
AI try-on, Zoom, supplier feeds, payment, shipping
Core commerce flow
Filters, AI relevance
On-device + cloud model
HST, shipping, promo
Supplier auto-selection
Auto consultation suggestion
step 1
→
step 2
→
step 3
Rules engine
Deterministic policies evaluated at every order, session and supplier sync.
Base × supplier markup × promo × HST(region)
Province + prescription → coverage band 0–80%
Nearest verified supplier with stock ≥ qty
Auto-book consultation 7 days post-delivery
AI try-on pipeline
A vision pipeline that segments hair, fits the chosen wig and renders a photo-real composite.
98-point detector, on-device
U-Net mask + scalp baseline
Cap-aware warp + albedo match
Latency p95 1.4s · cached per session
Data model
User
- · id
- · role
- · phone
- · city
Product
- · id
- · supplierId
- · price
- · stock
- · rating
Order
- · id
- · customerId
- · items[]
- · total
- · status
Consultation
- · id
- · customerId
- · consultantId
- · date
- · zoomUrl
Supplier
- · id
- · city
- · fulfillment%
- · status
CareNote
- · id
- · customerId
- · author
- · body
- · tags[]
Service-level objectives
API uptime
99.95%
Checkout p95
320ms
Try-on p95
1.4s
Support reply
< 2h
