I wasn't aware there was any other way to do this than Navier-Stokes in 3d? I am an algebra guy so I have no idea how PDEs work nor do I know how this kind of stuff is implemented, but to do an optimization where the algorithm figures out a shape that works would be like running the model millions of times. If you made changes and then gave the data to someone else to modify, that would be the same as what people are doing now wouldn't it?
Seems like some of the millions going to wind tunnels should be funnelled towards research grants for figuring out a better implementation of the numerical solution to Navier-Stokes! 10 million dollars is good for like, 20 postdocs + faculty advising for 5 years.
Between Toyota, Honda, Ford, Ferrari, GM, etc. + Boeing, Lockheed, etc. I bet they wouldn't have any trouble coughing up at least several times that number, and considering the cost of wind tunnel testing, I bet it would pay off.