While machine learning is gradually gaining widespread acceptance in Geo-scientific modelling and Geo-scientific data processing, its application in the fields of data assimilation is still in its infancy. However, the potential of modern machine learning techniques to integrate and improve current data assimilation methodologies is clear and the published use cases are growing fast. This session will focus on gathering, presenting and discussing applications of machine learning to aid, enhance or even replace traditional geophysical data assimilation. We invite use case examples as well as methodological contributions at the intersection of the two disciplines. These may include the handling of non-Gaussian errors, detection and correction of model errors, forecasting of extreme events, lifting the curse of dimensionality, handling nonlinear problems, bridging scales and coupled systems, making data assimilation computationally more efficient and many more.