That's My Pup — Case Study

A customer uploads one phone photo of their dog and gets back a 3D-printed figurine of it. That meant building the brand, the storefront, and an AI pipeline that turns a single snapshot into a manufacturable model.

Results

One photo Input Required — Remaining angles are generated, reviewed, and assembled into a printable model
Interactive 3D Customer Preview — Shoppers turn their pet's model in the browser and approve it before anything is printed
Human-approved Quality Gate — An automated vision check advises; a person signs off on every model that ships
Brand to fulfillment Scope — Identity, storefront, AI pipeline and manufacturing hand-off, all built for launch

The Challenge

A new direct-to-consumer brand, launched from nothing. The identity, the storefront and the product pipeline all had to exist before the first order could.The promise is deliberately hard. A customer uploads one photo of their dog (bad lighting, awkward angle, a phone camera) and expects a physical figurine that looks like their dog. One photo is not enough geometry to build from. Every other angle has to be generated, and generated consistently, because a model built from angles that disagree produces a figurine with the wrong face. And unlike print-on-demand, a wrong result costs real manufacturing money and lands on a customer's doorstep.

The Solution

The storefront is a custom React and Supabase build with Shopify handling checkout. The product is the pipeline behind it.A photo comes in and the missing angles are generated one at a time. The order is right, then left, then back. Every angle that passes review joins the reference set handed to the next generation, alongside the original clean front view. That accumulation is what makes it work: generating all the views at once produces angles that quietly disagree with each other, while each view referencing every approved view before it stays faithful to the same animal. The approved set becomes a 3D model, which the customer sees in an interactive viewer in their browser and approves before anything is manufactured.Every model passes an automated vision check that scores it against the source photos, and then a person signs off. That human gate is policy. The pipeline spends real money and emails real customers, so nothing customer-facing moves on an automated verdict alone. Around it sits the rest of a working business: lifecycle email, order tracking through to the manufacturer, review capture with customer photos and video, and consent-gated analytics built to a stricter standard than the law currently asks of a store this size.