Clear and comprehensive view on an automotive system design, as well as its analysis/optimization, are the key research topics in the automotive domain. To tackle the first problem, modeling languages such as the EAST-ADL2 or the AUTOSAR metamodel, have been defined. The second issue has found its solution in the numerous analysis and optimization techniques. Nevertheless consideration of these two problems simultaneously is not the case. This has resulted in many tools that deal either with modeling or analysis and optimization. As a consequence it is difficult to embed these tools in the standard methodologies such as the EAST-ADL2/ AUTOSAR methodology. This especially applies to the higher abstraction layers as those defined by the EAST-ADL2. It is a consequence of a high dependability of analysis/optimization tools on the platform specific information. Therefore this paper presents UML based framework that combines the modeling of the automotive architectures and support for an optimization at the early stage of a system design. Optimization technique uses genetic algorithms and the early stage optimization goals concern performance, safety and cost. We show how this framework complies with the EAST-ADL2/AUTOSAR methodology and present models expressing functional but also non-functional concerns related to the optimization goals.