Internet of Things will not be only a new worldwide network interconnecting trillions of (smart) devices but, most importantly, a platform (system of systems) where to develop a new wave of (cyber-physical) services for humans and machines. In this context, in order to build IoT systems, the so-called IoT-Edge-Cloud continuum paradigm is having tremendous focus from the research community as well as from the industry. This paradigm can therefore be an enabler to push intelligence from the core of the network to its edge: from centralized data mining to embedded machine learning in tiny IoT devices to federated machine learning involving networks of edge devices. Moreover, methodologies are emerging to support analysis, design, implementation and evaluation of solutions involving mining and machine learning at the IoT edge. In this talk, we will focus on IoT from both the architectural and machine learning at the edge perspectives. Finally, some use cases will be discussed related to mobile edge computing, ambient assisting living environments, and intelligent transportation systems.