Modelling multiphase flows involves not only the conservation laws of individual phases but also the way in which they interact with each other. This increases the complexity of modelling compared to single-phase flows. Two approaches can be used to understand multiphase flows. On the one hand, it can be purely data-driven, with heuristic models arising out of raw data, commonly referred to as black-box modelling. On the other hand, conservation equations derived from physics-based modelling can be solved to understand the flow evolution, referred to as white-box modelling. Both these approaches have their limitations. Black-box models can provide accurate predictions within the range of the data on which they are trained, but their predictive capability may deteriorate when applied to flow conditions outside the training dataset and they do not explicitly enforce the underlying physical laws. White-box models, while providing a physically consistent framework, require closure relations for unresolved multiphase interactions, such as interphase momentum and heat transfer, which are often difficult to model accurately across different flow regimes. The objective of this project is to develop a grey-box modelling framework that combines the predictive capability of data-driven methods with the physical constraints of conservation laws. In particular, machine-learning models will be used to learn unresolved closure relations from high-fidelity numerical simulations, while the resulting models will be embedded within the governing multiphase-flow equations and evaluated through a posteriori simulations.