The diversity of terminals used by the user to access resources (Personal Digital Assistant, mobile phone, etc.) using several types of networks (wireless, local, etc.) generates a growing need to adapt services dynamically to the user's context. In this article, we present our architecture that aims at adapting content and presentation of services to the user's context. We realize the content adaptation based on a new data modeling methodology. This methodology aims to take into account the different structures associated with the same data allowing a data representation in several viewpoints. The presentation adaptation is based on an automatic generation process of the interfaces service code. The context in our architecture is presented by a generic model for the user and the service. The adaptation process is detailed based on an e-learning scenario.