Offline on-device AI assistantPersonal systems project
Project Charlie
An Android assistant that runs entirely on the phone. Speech-to-text and a quantized language model, both native, with zero cloud API calls — built in Kotlin over a C++ JNI bridge.
- whisper.cpp for speech and a quantized Qwen2.5-0.5B through llama.cpp, fully offline
- Sequential load → infer → unload pipeline holds peak memory at 650–800MB inside a 2GB budget
- Native JNI bridge wiring two C++ inference engines into Jetpack Compose, with Room/SQLite for on-device memory
- Kotlin
- C++ (NDK/JNI)
- whisper.cpp
- llama.cpp
- Jetpack Compose
- Room/SQLite
github.com/vinaytz/ProjectCharlie
Built natively — screenshots coming soon
