While mobile computing devices have gained importance in recent years, thanks to their increased functionality and affordability, their diversity in hardware and software characteristics remains an obstacle to the provision of a common user experience across these heterogeneous systems. Furthermore, current user interfaces (UI's) running on these devices are developed with the assumption that the context in which they will be used is constant over time, including the profile of the user, the platform on which the user carries her tasks, and the environment in which the interaction takes place. Traditional software design techniques are not adequate to support multiple contexts, since they require a partial or complete re-write of the user interface to match the characteristics of each device and the software running on it. This in turn necessitates additional time and resources. This situation is further complicated by the need of the user interface to accommodate varying contexts such as the user profile and environmental conditions. Consequently, there is a need to develop a new software methodology that is flexible enough to take into account contextual aspects early on the design phase, and enables the separation of the interaction aspects from implementation concerns. One way to effectively achieve this is to use abstraction. To this end, we propose a model-based approach for adaptive user interface design and generation for resource-constrained devices. This approach consists in using a selected number of models which describe the UI at different levels of abstraction with specific terms and syntax. By focusing on the interaction aspects, it becomes possible to eliminate the need to rely on context-specific interaction capabilities, such as hardware and software, in the early stages of user interface development. It also becomes possible to enable context-adaptation at runtime, after the UI models have been defined. These models are specified using standard-based notations, and are then transformed and mapped to result in adaptive and adaptable user interfaces. A methodology to translate user requirements into a set of UI models is also introduced. Finally, a demonstrator application has been implemented to provide a practical example of how to apply these concepts and use semi-automatic processes developed as part of our research work. Key words: Context-aware User Interfaces, HCI, Model-based User Interface Development, UML, XML.