Kotlin Mullets - Business in the Front, Party in the Back
summary
One codebase builds Android UI, a shared common module and a Kotlin server (Micronaut) so client and server share a single data class for images and labels; on‑device inference uses ML Kit and TensorFlow Lite (custom digit model ≈5MB) while cloud labeling uses Firebase / Cloud ML and Cloud Run containers. Accelerometer traces get kriging interpolation on the server to render bitmaps for cloud ML; on‑device labels are often vague but cloud returns >90% confidences for faces and a digit demo hits 100% for a “3.” UI blocking from a synchronous TFLite interpreter is solved with Kotlin coroutines.
AI-written summary · about
chapters (9)
- 0:00Project introduction
- 1:45Android drawing mechanics
- 6:20Machine learning basics
- 13:31Custom ML digit classifier
- 18:41Full-stack integration
- 25:15Cloud-based inference
- 28:38Sensor data processing
- 33:24Containerized deployment
- 37:00Development lessons
related talks
51mKotlin Mullets - Business in the Front, Party in the Back
46mKotlin Mullets
32mKotlin Multiplatform in Action
41mKotlin for the JVM and beyond
15mKotlin Multiplatform Contest Winners
26mKotlin Multiplatform in an MVVM World
46mUnlocking Scalability: Power Up Your Native App with KMM
38mWeb and Android United- Kotlin-Powered Cross-Platform Excellence