Snow covers 40 % of Earth’s northern hemisphere each year, playing an important role in regulating Earth’s energy balance through its reflective properties and serving as the primary source of freshwater for one-sixth of the world’s population. Remote sensing is a key technology for observing snow across large areas. Given the complexity of retrieving the snow signal from space, the snow community is increasingly using machine learning (ML) and deep learning (DL) approaches to derive and improve current spaceborne snow measurements. Here we summarize the research trends and applications of ML for snow retrievals, identifying best practices and the future direction of ML applications for multiple seasonal snow-related properties, in particular snow water equivalent (a measurement of the amount of water contained in the snow), snow depth, density, cover, and wet/dry conditions. Across our review, we find that most studies focus on extrapolating to new areas and time periods, with improved prediction after combining multiple sensors and referencing high-fidelity reference data. Gaps remain in generalizing across spatial domains, combining sensors at different spatial and temporal resolutions, and improving the ability of newer model architectures, such as transformers and foundation models, to support snow-related applications. ML/DL for snow remote sensing is an effective tool for integrating information across sensors, and we recommend the snow community continue to identify further use, thereby advancing towards more consistent and complete global snow monitoring that includes a range of snow types and climate conditions.