Founding Engineer

Havenwise helps heat pump owners get more out of the system they already have. I joined as the founding engineer, which in practice has meant owning the backend, the mobile app and the AI layer, and deciding in what order each of them needed to exist.

Consolidating the backend into a platform

The product's logic was spread across more than ten Google Cloud Functions, each deployed and versioned on its own. Shared behaviour had been copied between them rather than factored out, so a change to one rule meant finding every place it had been duplicated — and missing one was an outage waiting to happen.

I established the core backend API platform and pulled that logic into it. The result is a codebase where a change lands in one place: less duplication, less operational risk, and a service that a second engineer can actually be onboarded onto.

Replacing the no-code app

The customer-facing app had been built on a no-code platform, which was the right call to prove the idea and the wrong one to build a business on. I built the mobile application from scratch in Expo and React Native, giving us an in-house product we could extend at our own pace instead of within someone else's constraints.

Payments and monetisation

I led the implementation of the payments workflow — the piece that turned Havenwise from a product people used into a product people paid for. It took the paid share of the user base to 70%.

AI where it actually pays for itself

Two pieces of AI work, both aimed at problems we could measure.

The first is an MCP-based integration with Intercom Fin AI that helps diagnose heat pump issues directly in the support conversation. Most inbound support is the same handful of diagnosable faults; the target is to cut support requests by up to 80%.

The second is Havenwise AI, a conversational interface that lets people interact with and control their heat pump in plain language rather than through a settings screen. That one is still being built.