Caressa Arias
Say hello →

Federal agency mobile app with on-device photo validation

Senior Mobile Developer (Public Trust) · Excella · 2021–2024

Problem

A U.S. federal agency’s public-facing mobile application (iOS + Android) asks users to submit photographs as part of a formal process, and those photographs have to meet clarity and compliance requirements before they are accepted. A blurry or non-conforming photo that only fails after upload wastes the user’s time and the agency’s review capacity. The application was also built and shipped inside government security, deployment, and review constraints, so neither the architecture nor the release process had much room to improvise.

My role

I led the React Native and TypeScript architecture for the application as the sole mobile engineer, delivering secure, scalable features under federal compliance requirements. I worked in a SAFe-style agile environment, participating in PI planning, estimation, prioritization, and story breakdown, and coordinating with QA through development, testing, and release readiness. I also contributed mobile architecture strategy to contract proposals and recompete submissions, aligning the technical approach with client and agency requirements.

What I owned

Cross-platform parity between iOS and Android, and the custom native modules — Swift on iOS, Java on Android — that extended React Native where the framework alone was not enough. Google ML Kit integration to validate image clarity and enforce photo compliance requirements before submission. OAuth authentication and Secure Store for sensitive user data. Test suites in Jest and React Testing Library, maintaining 90%+ coverage on core features. The full release pipeline: App Store and Google Play deployments automated with Fastlane inside a regulated release process. And translating complex Adobe XD designs into responsive, accessible UI.

Key decision

The photo validation runs on the device rather than on the server. That keeps sensitive images off the network until they’ve already passed the clarity check, and it lets a user retake a photo immediately instead of waiting on a round trip to learn it was rejected. The cost is that on-device ML meant writing and maintaining two parallel native modules, in two languages, with two sets of platform quirks — a permanent maintenance surface I took on deliberately in exchange for the privacy and latency properties.

Result

I shipped secure cross-platform features that met federal compliance requirements, and owned the regulated release pipeline for both app stores end to end. High test coverage on core features kept the codebase safe to change under a single-engineer maintenance model. The mobile architecture work I contributed to proposals and recompetes supported contract awards and renewals.

Built with: React Native · TypeScript · Swift · Java · Google ML Kit · Fastlane · OAuth · Jest