Adaptive learning and performance support systems are considered superior to human trainers in monitoring user behavior and providing personalised support and feedback. However, the process of their development is complex and expensive, and they are typically not transferable to domains other than those for which they were intended (e.g. from maths to reading). Moreover, their adoption is barred by a lack of didactical integration. LEAPS will tackle these challenges by investigating and implementing an approach that combines behavioral data with learning analytics, machine learning, and learning dashboards. This approach will be evaluated in three demonstrators for four very specific populations: training of number sense in pre-school children, training of initial reading for primary school children, compensatory support in writing for people with dyslexia, and on-the-job development of written language skills. The interdisciplinary consortium brings together expertise in statistics, computer science, educational sciences, cognitive science, human-computer interaction design, software engineering, and through knowledge of user populations in the markets of school education and professional training.