Selected work

Projects, with the numbers attached

Three engagements we're allowed to talk about, described the way we'd describe them to another engineer — including what was actually hard.

0Fewer escaped defects
0From a 4-hour suite
0Crash-free sessions
0Unplanned outages
Freight & supply chain

A twelve-year-old dispatch system, replaced without stopping the trucks

The problem. Every quote went through a desktop application that only ran on Windows 7 machines in one building. Nobody dared change it. Growth was capped by the number of people who could sit at those desks.

What we did. A strangler-fig migration: we put an API in front of the legacy database, moved one workflow at a time to a new web platform, and ran both systems side by side until the old one had nothing left to do. Fourteen months, no cutover weekend.

  • .NET
  • React
  • Azure
  • SQL Server
  • Terraform

Meridian Logistics · Outcomes

Quote turnaround down from 40 minutes to under 3
Dispatch team working remotely for the first time
Zero unplanned outages across the entire migration
Legacy licence and hardware costs eliminated
Digital health

A patient app that stopped crashing and started retaining

The problem. A React Native app with a 2.8-star rating, a 9% crash rate and a release process that took three days of manual regression testing before anyone would press the button.

What we did. We rebuilt the data layer around offline-first sync, moved the two most fragile screens to native Swift and Kotlin, and built an Appium suite covering the twelve journeys that mattered. Releases went from three days to forty minutes.

  • Swift
  • Kotlin
  • React Native
  • Appium
  • Firebase

Kestrel Health · Outcomes

Crash-free sessions from 91% to 99.7%
Store rating 2.8 → 4.7 within one release cycle
Release cycle from 3 days to 40 minutes
30-day retention up 34%
Regulated fintech

40,000 automated tests, and a Friday deploy that nobody dreads

The problem. A regulated lending platform with 200 engineers, a four-hour test suite that failed randomly, and a release calendar that had shrunk to once a month because nobody trusted the pipeline.

What we did. We triaged the flaky tests ruthlessly, deleted the ones testing nothing, parallelised the rest across ephemeral environments, and added contract testing between the twelve services that kept breaking each other.

  • Playwright
  • Pact
  • k6
  • Kubernetes
  • GitHub Actions

Northfold Financial · Outcomes

Suite runtime 4 hours → 11 minutes
Flake rate 18% → under 0.4%
Monthly releases → 30+ per week
Production defect escape rate down 71%

Your project could be the fourth one on this page

Tell us what you're building and we'll tell you honestly whether we're the right people for it.