In an effort to improve and automate the fluid dynamic design of rotary blood pumps, a coupled computational fluid dynamics (CFD) shape optimization methodology has been developed and implemented. This program couples a finite element flow simulation with a gradient-based optimization routine to modify automatically the shape of an initial candidate blood path, according to a variety of desired fluid dynamic criteria, including shear stress, vorticity/circulation, and viscous dissipation. Preliminary results have led to both intuitive and nonintuitive transformations of the initial blood flow paths for both internal and external flows. This application of computer design optimization offers the ability to explore a much broader design space much more efficiently than would be possible with traditional parametric methods. It is believed that this computer tool can assist developers of rotary blood pumps in designing blood-wetted components that minimize thrombosis and hemolysis while simultaneously providing maximum flow performance.