Fine-Tuning Whisper.cpp for On-Device Speech-to-Text in KMP

Deep technical walkthrough of integrating Whisper.cpp into a Kotlin Multiplatform project using expect/actual declarations for platform-specific audio capture (AudioRecord on Android, AVAudioEngine on iOS), comparing int8 vs int4 quantization tradeoffs on mobile hardware, implementing a sliding-window chunked inference pipeline that keeps memory under 200MB, and building a Kotlin coroutine-based streaming architecture that feeds audio frames to the model while rendering partial transcripts at 60fps — all running entirely on-device with zero API costs.