This session describes the architecture of the Custom Vision service to build ML image classifiers, to bulk label images with a REST API, and to train a model and use it online and offline in a mobile app. Custom Vision provides developers with the possibility to personalize the image detection and analysis service of Azure Cognitive Service, to tailor the data model to specific use cases. Trained model can be utilised also in environments with no live connectivity to the cloud, offering the possibility to deliver intelligence at the edge. The session describes also the continuous learning of the ML classifier based on newly collected images, and how to create a time series of images persisted in Azure storage using serverless functions.