Dongmin, Yu

• Drove React Native client development for a rewards app on a two-person team: over 82 days I authored 77% of the repository's non-merge commits (1,054 of 1,369) and 144 pull requests, and left behind 105 test files, 115 technical documents, and 43 CI automation files.
• Single-handedly completed the React Native New Architecture (TurboModule/Fabric) migration and the RN 0.81 + Node 22 LTS upgrade, and designed a deploy-tracking tag convention while running 15+ store releases and 19+ OTA deployments.
• Solely designed and built three native libraries — on-device receipt OCR (ML Kit, VisionKit + Vision), a reward-ad SDK, and an offerwall SDK — moving receipt-recognition preprocessing from the server to the client. Also implemented Naver OAuth 2.0 login and a module wrapping Google Play In-App Review and Apple StoreKit `requestReview` behind a single API, in Swift/Objective-C and Kotlin.
• Unified four analytics SDKs (Amplitude, AirBridge, Notifly, Firebase) behind a single `Analytics.track()` interface, migrated 30+ files off direct SDK calls, and pinned the event taxonomy to a codegen contract that gates schema drift.
• Owned and filed 220 Jira issues, 139 of which shipped, and resolved 14 user-reported incidents. Diagnosed three separate root causes behind unpaid ad rewards and then designed a server-to-server callback structure, and — at 96.62% crash-free users — introduced a crash-ticketing process and an error-reporting abstraction.

• Led the 0-to-1 launch, operational stabilization, and continuous improvement of a Flutter (iOS/Android) app for a pet behavior/personality-type assessment and expert consultation platform.
• Resolved payment/policy issues (Apple review/payment policy) and built three PG payment integrations on PortOne, laying the groundwork for launch and B2B/B2G expansion.
• Built an observability system based on crash, log, and behavioral (funnel/retention) data to drive product improvements.
• Structurally managed state, model, and routing complexity using Riverpod, Freezed, and go_router.
• Achieved 85%+ positive responses for UI/UX in a user satisfaction survey.
• Led Apple DevRel meetings and resolved policy issues by arguing the service's characteristics (domestic-only, offline brokerage).
• Standardized the payment flow on PortOne and developed and published the `portone_flutter_v2` package.
• Led contracting and integration for three payment gateways (Toss Payments, Toss Pay, Kakao Pay).
• Built unified logging on Sentry + Crashlytics, reducing the time needed to identify the root cause of incidents.
• Linked analytics, session recording, and event logs by user identifier to lay the groundwork for funnel/retention analysis.
• Reliably handled complex entry/return flows with URL-based routing.
• Standardized the install/deep-link experience across iOS and Android using Universal App Links.
• Built a non-developer editing/publishing workflow that reduced content operations time.

Wegooli is a mobility startup that has developed a car-sharing platform for public housing residents, including Bucheon's Ministry of Land smart-city project, among others.

• Led a rapid response to the existing app's Android SDK issues and the launch of a new Flutter app, improving overall app quality and performance.
• Rapidly built a componentized UI on a Figma design system, cutting development time by more than 30%.

Gbike is a mobility startup that operates the electric kickboard sharing service 'GCOO' across 50 cities domestically and internationally.

• Directly developed a React Native module that subscribes to real-time gait data computed from each device's supported sensors.
• Gained performance-optimization experience by directly using modern native module architectures and performance-monitoring tools such as Turbo Module, Hermes, and Flipper.
• Developed a native module that combines iOS's CMPedometer and Android's SensorManager to process step-count data in a unified way across platforms.
• Improved measurement accuracy by applying the AOSP step-detection algorithm to low-level accelerometer data, to support devices down to Android API 19.
• Built on a React Native Turbo Module-based architecture to reduce the JS-Native bridge bottleneck and improve response latency.
• Monitored and optimized memory usage and runtime behavior using Hermes/Flipper-based performance tracing.

Bside Korea is a shareholder-activism platform that lets minority shareholders in invested companies act together through shareholder verification, consolidation, and electronic signatures.

• Took part in overall development of the Flutter-based app and the RedWoodJS-based web app, understanding and implementing requirements on both the backend and mobile sides.
• The Flutter app was designed around proxy-voting delegation during the season, and automated shareholding verification and lounge-gathering features in the off-season.
• Implemented advanced data-integration features via Kotlin/Swift-based native functionality and a GraphQL API.
• Over 1,500 users participated in the actual proxy-voting delegation process, recording an average user feedback satisfaction score of 4.5.
• Using Flutter, built a service that collects electronically-signed proxy votes from minority shareholders for a specific stock during a campaign and delivers their consolidated opinions to the asset manager through the shareholders' meeting.
• Built the main components using Flutter's Material UI and official packages, and implemented ID card photo capture, e-signature storage, and voting for proxy delegation.
• Wrote native bridge code in Kotlin (Android) and Swift (iOS) to integrate with external scraping and the accredited-certificate (공인인증서) module.

A federated Flutter plugin providing receipt image capture, crop, orientation normalization, JPEG compression, EXIF extraction, and on-device OCR on iOS and Android (pub.dev: https://pub.dev/packages/flutter_receipt_scanner).

• Designed as a federated plugin split across an app-facing package, a platform interface, and the iOS/Android implementations, with the platform message contract generated by Pigeon.
• Implemented both the camera and gallery paths — VisionKit document scanner with Vision OCR on iOS, the GMS document scanner with ML Kit OCR on Android — with a Dart-side OCR-floor gate that rejects captures below a recognition-quality threshold.
• Drew the boundary so the native layer owns image primitives only, leaving store/amount/date parsing, upload, and cloud OCR to the consuming app. Solely responsible for design, implementation, and pub.dev publishing.

A React Native receipt scanner that wraps ML Kit (Android) and VisionKit + Vision (iOS) behind a single API (npm: https://www.npmjs.com/package/react-native-receipt-scanner).

• Built exclusively for the New Architecture — calls go through JSI on iOS and the codegen-generated spec on Android, with no legacy bridge path. Results are returned as `file://` URIs only, never base64.
• Provides camera and gallery scanning, interactive perspective-crop, JPEG compression, EXIF extraction, and on-device OCR for Korean and Latin script — with no network call and no external API key.
• Targeted the gap left by existing document-scanner libraries — on-device OCR and TurboModule support on both platforms — and solely handled design, native implementation, and npm publishing.

Merry is a script manager for Dart/Flutter projects that lets you define long, hard-to-remember scripts under short names and run them (pub.dev: https://pub.dev/packages/merry).

• Started from derry but added many features, and also built the vscode-merry VS Code extension to integrate into the existing ecosystem.
• vscode-merry shows script definitions in a browsable UI inside the VS Code Explorer, so they can be discovered, checked, and run without switching between YAML files and the terminal (published on the VS Code Marketplace).
• Solely designed, implemented, and published both the Dart package (merry) and the VS Code extension (vscode-merry).

A React Native library that subscribes to the user's step count via Android's StepCounter sensor and iOS's Core Motion (npm: `react-native-step-counter-newarch`).

• With 44 GitHub stars and 21 forks, this is the most widely adopted of my personal open-source projects (as of July 2026).
• Broadened device coverage with an accelerometer fallback for Android devices that lack a step-counter sensor, and uses Core Motion's CMPedometer on iOS.
• Built on a TurboModule architecture with a dedicated documentation site, and solely handled design, native implementation, and npm publishing.