Geometric deep learning (GDL) provides a unifying framework for today's seemingly diverse deep neural networks used in learning from images, graphs, and sets. In this mini-course, we give an overview of the geometric deep learning framework and discuss the role of computational optimal transport in GDL as a fundamental tool to manipulate and compare probability distributions. In recent years, the Wasserstein distances arising from the optimal transport problem have been particularly interesting to the machine learning community. Our mini-course is structured as follows: (Part A) An overview of geometric deep learning, (Part B) A crash course on Optimal Transport, (Part C) Wasserstein embeddings, and (Part D) Wasserstein embeddings in the geometric deep learning. The mini-course will be centered on our recent publications on the topic at ICLR 2020, NeurIPS 2019, and NeurIPS 2021.