A couple of production problems told the way I'd talk through them in an interview: the constraint, the approach, and what actually shipped.
Replacing ML Kit with Apple Vision for iOS face detection
Replaced the iOS ML Kit dependency in react-native-vision-camera-face-detector with Apple Vision to resolve the arm64 iOS Simulator limitation while keeping Android on ML Kit.
Migrating a production app to React Native's New Architecture
Took a large, live React Native app off the legacy bridge and onto Fabric + Hermes — while it kept serving 100k+ daily users.
Shipping reliable multi-provider payments in a mobile app
Integrated multiple payment providers behind one dependable checkout, with native iOS handling where it mattered.
A refer-and-earn flow that actually attributes, with deep linking
Built referral sharing and deep linking so a shared link opens the right screen — or the store — and the reward is correctly attributed to the referrer.
Real-time chat & in-app support that lifted conversion 60%
Built an end-to-end real-time chat and in-app customer support system — backend to frontend — that kept students and tutors talking and lifted conversion by ~60%.