Machine Learning Systems

Production Android AI with ExecuTorch 1.0 - Deploy PyTorch models to mobile with NPU acceleration and 50KB footprint

LLM inference on mobile via Capacitor — run quantized GGUF models on-device

GGUF/llama.cpp inference and chat runtime for Dust — iOS/macOS with Metal GPU

📱 Optimized ML for edge devices. Showcasing efficient model deployment, GPU-CPU memory transfer optimization, and real-world edge AI applications. 🤖

Device Unified Serving Toolkit — central hub for the Dust on-device ML ecosystem

On-device text embedding generation for iOS and Android via Capacitor

Model download and serving orchestration for Dust — Capacitor bridge

Android ML model server — download management, session caching, accelerator probing

Standalone ONNX runtime session management and preprocessing for Dust — iOS/macOS

DustCore contract types and interfaces for on-device ML — pure Kotlin/JVM

 INT8 quantization of MobileNetV2 for learning and production-oriented iOS mobile inference.

Claude Code skill for Google LiteRT - on-device AI/ML deployment framework

Core Capacitor bridge for the Dust on-device ML framework — iOS and Android

A lightweight, mobile-optimized Neural Machine Translation (NMT) framework in PyTorch. LingoLite features a modern transformer architecture with state-of-the-art optimizations for efficient multilingual translation on resource-constrained devices.

Android ONNX runtime session management and preprocessing for Dust

Standalone model server business logic for iOS — download, caching, accelerator probing

Standalone tokenizers and embedding runtime primitives for Dust — iOS/macOS

DustCore contract types and protocols for on-device ML — pure Swift

Android GGUF/llama.cpp inference and chat runtime with JNI bindings