Tiny Machine Learning Systems: From Hardware to Applications

The growing demand for intelligent devices at the edge calls for machine learning systems that are not only accurate but also resource efficient. This talk presents recent advancements in tiny machine learning: a paradigm that brings AI capabilities to low-power, embedded platforms. We will explore how "emerging computing" models (such as stochastic and hyperdimensional computing) can be exploited to design lightweight, always-on learning systems. By co-designing hardware and software, we can push the boundaries of what is possible in constrained environments, enabling applications ranging from real-time health monitoring to secure embedded systems.