Vehicle Operations Self-Management PWA

Problem

The self-management path still assumed a staffed counter or a brittle mobile site. Owners could not complete the core loop from a phone. Adjacent ops tools (back-office, picker, and client apps) had drifted out of version together, so a change in one surface did not survive in the others.

Constraints

The product had to ship as a PWA, not a store app. Existing identity and policy systems stayed the source of truth; a rewrite of those systems was out of scope. Test coverage had to rise with the rebuild, not after it. This write-up does not describe the insurer, the vehicle catalog, or any production UI.

Architecture

Version 2.0 of the owner-facing surface was built from scratch in React as a PWA: installable, offline-tolerant for the last-known session, and driven by a typed API client. A shared component and contract layer kept the back-office, picker, and client apps on the same version line. TDD covered roughly three quarters of the new work so regressions in the owner loop could not hide in the ops tools. A separate Electron kiosk handled in-person client interaction without sharing the PWA's routing.

Role

As frontend technical leader I owned the PWA rebuild, the shared contract layer across the three ops surfaces, and the test discipline that made that sharing safe. I also shipped the Electron kiosk as a distinct in-person path.

Outcomes

• Owners could complete the self-management loop from a browser without a native install.
• Back-office, picker, and client apps moved as one version instead of three forks.
• TDD covering roughly 75% of the new work made the shared contracts enforceable.

Stack