Abstract
We describe an attempt of bridging the gap between web-based learning and agents capable of learning from experience. The emphasis is made on the interaction between two core fields, namely agents and Unified Modeling Language (UML) as a standard for Object Oriented Design. The tangible results remain the integration of agents for elearning based on machine learning methods such as entropy-based Decision Tree Learning, AdaBoost and First Order Inductive Learning (FOIL). As a special case of soft computing methods, fuzzy agents are used for profile personalization. Prospectively, much effort is still required to meet the actual challenges so as to scale up to real-life problems of any significant complexity.