Separation of wall bounded flows for Navy relevant geometries is a highly complex phenomenon. Due to the shape of underwater vehicles (submarines, torpedoes, low aspect ratio lifting or control surfaces) separation is often three-dimensional (3D). Because of the relatively high Reynolds numbers involved, separation is always associated with considerable unsteadiness. This unsteadiness is caused by large coherent structures that are a consequence of hydrodynamic instability mechanisms of the mean flow. The combination of threedimensionality and unsteadiness results in a highly complex time-dependent topology of the separated region. We are simulating 3D separation bubbles in internal (diffuser) and external flows. The diffuser flow simulations are conducted in collaboration with experiments at Stanford University by J. Eaton. The flow is turbulent and difficult to access by direct numerical simulations (DNS). The objective here is to develop, validate, and advance appropriate hybrid turbulence modeling capabilities that will lower the computational expense associated with such simulations. In a combined numerical/experimental effort we are also studying laminar separation bubbles in external flows. For these simulations we are employing highly resolved DNS. The objective here is to obtain high-fidelity flow data that will be analyzed to obtain a deeper understanding of the various physical mechanisms governing separation, transition, and reattachment of 3D bubbles. Ultimately, such understanding may pave the way for the development of effective and efficient flow control strategies for preventing separation in practical applications.