Model-based Machine Learning (MBML) is a popular approach for creating bespoke solutions tailored to real-world applications. The core idea at the heart of model-based machine learning is that all the assumptions about the problem domain are made explicit in the form of a model. This talk will focus on the design and application of MBML methods to industrial applications inspired by crowd sourcing domains where a key challenge is to retrieve high-quality data from human judges with unknown accuracy and biases. It will demonstrate how these methods are able to improve the quality of crowd sourced information by simultaneously learning the unknown judges’ reliability and the aggregated labels. I will also discuss future directions and new research challenges arising from recommender systems and to privacy-aware learning systems.