Semantic Location through On-Device Smartphone LLMs and WiFi Sensing

What is it about?

We’ve developed an Android application that leverages WiFi SSIDs to infer environmental context, powered by a fine-tuned, on-device LLM for real-time location insights. Our solution runs entirely on-device, offline and in a privacy-preserving manner. By leveraging context-aware intelligence, we can deliver personalized, privacy-preserving context that enhances user experience in mobile health apps and environment-aware applications.

Featured Image

Photo by Shuvro Mojumder on Unsplash

Photo by Shuvro Mojumder on Unsplash

Why is it important?

Focusing on on-device semantic location inference without dependence on GPS, internet connectivity, or cloud services presents a critical shift as privacy and real-time responsiveness become central concerns in mobile computing. While most location-based systems prioritize coordinates or rely on external APIs, our approach brings contextual intelligence directly to the edge, empowering devices to interpret their surroundings through ambient signals like Wi-Fi SSIDs. This not only expands the possibilities for offline, privacy-aware applications, but also opens new avenues for context-adaptive mobile agents, health interventions, and assistive technologies that can react meaningfully to a user’s environment in real time. By demonstrating that this is possible even on older smartphones, our work helps push forward the accessibility and scalability of edge AI.

Perspectives