Cloud AI services have attracted tens of millions of paid subscribers, but the potential reach of AI-ready smartphones is considerably larger. Industry estimates place the active base at between 350 million and more than 500 million devices, with the installed footprint moving towards one billion as consumers upgrade their phones. What happens when AI moves beyond a remote cloud service and becomes something that runs directly on the devices people carry every day?

In this session, you will explore the capabilities and limitations of running small, purpose-built AI models locally. You will examine how constraints such as speed, privacy, and immediacy influence software architecture and user expectations, and how sub-billion-parameter models and edge embeddings can enable fast inference, offline execution, and local data processing without relying on cloud APIs.

The session considers why the most valuable AI solution is not necessarily the largest model available, but the model best suited to the task and the hardware on which it runs.

What You Will Learn

• Why the growing installed base of AI-ready smartphones creates new opportunities for on-device AI

• How sub-billion-parameter models and edge embeddings enable local AI capabilities

• How speed, privacy, offline operation, and hardware constraints influence application architecture

• How to evaluate the capabilities and limitations of small local models against cloud-based alternatives

Who Should Attend

Software Developers, Mobile Developers, AI Engineers, Software Architects, and technical practitioners interested in on-device AI, edge inference, and privacy-conscious application design.