Divy Arora

PocketMesh is an open-source Edge AI infrastructure project that transforms Android smartphones into portable AI servers. It enables users to run Large Language Models (LLMs) directly on their devices and expose AI inference through a local HTTP API, allowing laptops and other connected devices to access AI without relying on cloud services or an internet connection. Currently powered by Qwen 0.5B using llama.cpp and quantized GGUF models, PocketMesh is being developed to support Meta Llama models and a broader ecosystem of locally deployable AI models. Built with Kotlin, C++, JNI, and Jetpack Compose, PocketMesh aims to make AI more accessible, private, and decentralized by turning everyday smartphones into independent compute nodes. The long-term vision is to connect multiple smartphones into a collaborative AI mesh network where devices can discover each other, share computational capabilities, and distribute AI workloads over local networks. PocketMesh - Your phone is not just a client. It's infrastructure.