Brownfield AI SDK
An embeddable React Native SDK that adds AI local-guide features to a tourism client’s live apps. Integrated by their app agency without access to the source.
The challenge
Integrate an in-app AI travel guide into a tourism agency’s live iOS and Android apps: chat with voice, follow-up questions, and local recommendations.
I never had the source code. An outside agency maintained the apps, so my work had to drop in cleanly, without disrupting the live product. The guide also couldn’t be a separate screen. It had to plug into the app’s existing native features.
The hardest part was lifecycle. Cold start and warm start differ across iOS and Android, and users constantly move between the native app and the SDK surface. It had to behave correctly in every state, on both platforms.
My role
Owned the SDK solo: architecture, packaging, releases, debugging production issues during integration. Four milestones, provided designs, direct technical contact for client and agency.
What I delivered
- The AI guide: chat with voice input; inline map links and follow-up questions; narrated points of interest
- Integration into the existing features: open recommendations on the map, favorite places
- Plug-and-play SDK: XCFramework (SPM) and AAR (Maven). Consumed like any dependency.
- Showcase apps: rebuilt the client’s app shell to showcase integration end to end.
- Error and crash monitoring, opt-in analytics, clear integration & developer guides.
- Iterative releases: automated CI and changelogs.
Results
- Live since 2026 in an app with 100K+ downloads (App Store and Play Store).
- Reusable SDK architecture: for further brownfield SDK integrations.