PawVault

A friend introduced us. They had a problem, not a plan.

Two founders came to us through a mutual friend. They did not have wireframes or a spec or a half-finished app from another developer. They had noticed something and could not let it go.

Everybody has hundreds of photos of their pet. They sit on a phone, mixed in with screenshots and receipts, and there is no good way to share them or keep them meaning anything. Then the pet dies, and those photos become the only thing left. They still sit in the same camera roll, buried under everything else.

That was the whole brief. Photos of a life, with nowhere to live.

We were the first developers they talked to. That cuts both ways. Nobody had already made a mess we had to clean up, but it also meant every early decision was ours to get right, and they had no way to check us.

What we built

A Flutter app for iPhone with an AWS backend. Subscriptions run through RevenueCat. There is a gifting flow, so you can buy the app for someone who just lost a pet, built on consumable in-app purchases and Apple’s Grant Entitlement API. PostHog for analytics. Firebase for notifications.

Three features carry the product. LifeTale turns a pet’s life into a written story. Reflections is guided journaling, a set of prompts that walk someone through what the animal meant to them. ToonMemory turns photos into illustrations.

The work was scoped as a fixed statement of work with four milestones, paid on delivery rather than by the hour.

The argument we had

ToonMemory was the hardest conversation on the project.

AI image generation is not consistent. Run the same photo twice and you get two different results, and some of them are wrong in ways that are hard to predict. On most products that is a bug you file and fix later.

This is not most products. The photo is somebody’s dead dog. If the output is off, if the eyes are wrong or the face is not quite the right face, you have not shipped a glitch. You have handed a grieving person a picture of an animal that is almost their animal.

We pushed back on how the feature was framed and how much the app should promise. The founders had the better instinct on why the feature mattered emotionally. We had the better read on what the technology would actually do on a bad day. Both halves needed to be in the room.

The part that took longest

Store review. Android especially.

Google rejected the app repeatedly, for a list of reasons. Store policy is not one rule you break, it is a set of overlapping requirements about data handling, permissions, billing, and disclosure, and each rejection tells you about one of them at a time. Every round trip costs days.

This is the part founders never budget for. The build was done and the app still was not out. There is no clever engineering answer, only reading the actual policy, fixing the actual thing, and resubmitting.

Where it landed

Approved on the App Store in June 2026. There are real users on it now and the number keeps going up. Android is still in progress.