This work focuses on the development of a CAD-integrated shape optimization framework based on the FlowSimulator HPC ecosystem and the gradient-enabled API named FSMDAO that allows native support for, and interoperability, with OpenMDAO. The algorithmically differentiated CAD-kernel OpenCascade Technology (OCCT) [1] was integrated as a plugin into the MDAO framework environment. This plugin allows a direct simulation-link-to-CAD and offers exact forward/reverse mode differentiation of geometric sensitivities with respect to the CAD design parameters. Workflow-integration of the true CAD kernel is deemed to have a number of advantages over CAD-free approaches, among them: the optimization is based on the true and exact CAD model, and geometric constraints with sensitivity derivatives can be conveniently handled in the original CAD system. This way, one avoids redundant or sophisticated auxiliary shape descriptions like free-form deformation (FFD) or CAD-surrogates. Therefore, the optimal geometry is available in the original CAD parametrization. Nevertheless, CAD-based workflows call for a reliable mesh-to-CAD link and require integration of the CAD software into HPC workflows to leverage high-fidelity MDAO. The presented CAD-based shape optimization workflow consists of the following gradient-enabled components integrated via FSMDAO: (i) the CAD plugin FSOCCT that integrates the AD-supported OCCT into the FlowSimulator ecosystem, (ii) a mesh deformation plugin, which translates changes in the surface to the volume mesh and (iii) the CFD Software by ONERA, DLR and Airbus (CODA) [2]. Selected aerodynamic airfoil and aircraft/wing-body configurations of different complexity, mesh resolution and CAD models are studied in this work in order to evaluate the presented approach. We demonstrate and verify the automated and AD-based sensitivity analysis. Attention will be given to the CAD-based handling of constraints in the MDAO workflows. Moreover, computational efficiency aspects of the CAD-integrated framework approach in conjunction with high-fidelity simulations will be discussed in the presentation.