Abstract

EdgeMob is a mobile-native AI compute ecosystem that transforms billions of smartphones into a globally distributed inference network. Instead of relying on centralized GPU clouds that are expensive, privacy-limiting, and often inaccessible to smaller teams, EdgeMob enables developers to deploy and serve both open-source and custom models directly from mobile devices. With the EdgeMob app, a developer can load a model locally, expose it via an API Gateway, and test or integrate it into applications without any dependency on cloud AI providers.

At its core, EdgeMob combines a React Native mobile shell with Rust and C++ runtimes, integrating frameworks such as Candle and llama.cpp to deliver efficient on-device inference. For larger models, EdgeMob supports distributed execution by partitioning model layers across multiple devices, allowing collective inference that scales with the network. While service-level agreements (SLAs) are limited today, they improve as mobile hardware advances, positioning EdgeMob as the foundation for long-term, mobile-first AI infrastructure.

The platform is designed as an ecosystem for AI compute in mobile phones, supporting both Web2 and Web3 applications. Developers can leverage EdgeMob’s SDKs and libraries to integrate local or gateway-hosted inference into apps, while decentralized applications (dApps) can tap into EdgeMob as an AI backend for DeFi, DeSci, gaming, NFTs, or privacy-preserving background jobs. With offline capabilities, EdgeMob also supports critical use cases in disconnected environments such as field operations, military, or remote deployments.

EdgeMob’s roadmap extends beyond model inference. It is visionary in exposing custom Model Context Protocol (MCP) servers, enabling on-device or distributed model training, and orchestrating background batch AI compute jobs across the mobile network. These capabilities position EdgeMob not only as an inference engine but as a full-stack mobile AI infrastructure for next-generation applications.

The economic layer of EdgeMob is powered by the EGMO token on Solana. Node operators—individual users who contribute their mobile CPU and RAM—earn rewards for running inference and serving requests. Developers and dApps spend EGMO to access compute, route requests through the gateway, or scale workloads. This creates a self-sustaining ecosystem that aligns incentives between developers, mobile operators, and the wider community.

By unlocking the world’s largest untapped compute resource—smartphones—EdgeMob introduces a new category of Mobile AI Compute Infrastructure. It bridges the gap between cloud-scale AI and decentralized physical infrastructure networks (DePIN), delivering cost efficiency, privacy, and accessibility to the next generation of AI-powered applications.