Machine Learning has revolutionized modern societies by enabling the automatization of diverse products and business processes. Unfortunately, in most of the cases, data scientists build one machine learning model for each task, which results in repeated efforts for solving similar problems. In order to "transfer" know-how among similar tasks, serious efforts are spent on jointly optimizing models for a set of tasks, or transferring models from a task to another. In that manner, the generalization quality of the predictions is boosted, especially on problems with limited supervised instances. During this seminar, students will have the opportunity to study the state-of-the-art research literature on multitask and transfer learning across diverse domains. The outcome of the seminar is to enable students to understand, analyze and positively criticize scientific research.