Projects

Selected work in Android, Edge AI, and offline-first engineering.

I-ScanTea (Offline Disease Detection)

Built a fully offline image-classification engine using TensorFlow Lite, optimized for low-end devices with <200ms inference. Designed a custom preprocessing pipeline to resolve prediction drift and achieved 99% accuracy.

Kotlin • Jetpack Compose • TFLite • Coroutines • Custom Views

Scanberry (Cloud-Integrated Quality Classifier)

Developed a scalable MVVM architecture for real-time image upload and classification. Implemented coroutine-based async flows to prevent UI blocking and secured HKI copyright for the system design.

Kotlin • Compose • Retrofit • MVVM • Coroutines

Talentid (Production Internship Feature Development)

Refactored core features for offline-first usage and stabilized the legacy codebase by resolving crash loops and memory leaks. Improved app reliability for users in unstable network environments.

Java/XML • Kotlin • Android Studio • SharedPreferences