This project's vision is to unify sensors and signals from diverse domains, including physical, health, weather, and more, into a common language.
Our first proof-of-concept was a physical activity assistant, SensorFit AI, that utilizes real-time movement sensor data from a smartwatch to give personalized feedback and suggestions. It has memory, planning, and reasoning features, enabling it to give you personalized feedback on your skill level, mistakes in your form, tips to improve, and more.
We are currently investigating a common framework and characteristics to analyze for signals and collecting data from new sensors. In the future, we plan to release a theory-inspired sensor language framework with a corresponding benchmark of expert-annotated data from diverse sensors and tasks, as well as a first-of-its-kind sensor foundation model for many tasks.