A novel geometry modeling technique is defined for the optimization of pressure recovery through a twodimensional subsonic diffuser based on that of a Formula One race car airbox. The airbox design procedure involves considering the expansion of the flow entering the airbox coupled with a bend through 90 deg. Both of these features are discussed separately in terms of parameterization approaches before the most suitable techniques are united in the final optimization study producing an airbox harboring strong local geometric control and with a set of design variables compact enough to retain optimizer efficiency. A global Krig response surface model is employed to support our optimization studies, comprising design of experiments and updates based upon the expected improvement to the objective function followed by a local exploration in a reduced area of the design space. We find that we can efficiently converge to an optimum airbox design. The geometry modeling technique discussed allows for potentially radical designs with high pressure recoveries. Nomenclature A = two-dimensional diffuser width, m a = Hicks-Henne bump amplitude, m C p = static pressure coefficient E = two-dimensional diffuser total expansion ratio, A e =A in Ex = local expansion ratio, Ax=A in f = objective function k = number of dimensions N = diffuser axial length, m n = number of observed responses p = mass-averaged static pressure, Pa R = correlation between two sample points r = vector of correlations T = dimensionless width factor of Hicks-Henne bump U = mass-averaged inlet velocity, m=s x = untried point x p = normalized distance of Hicks-Henne bump peak along duct centerline from duct entry, x p 2 0; 1 = wall-contouring parameter ij = Kronecker delta where ij 1 if i j and ij 0 if i ≠ j , p, = Kriging hyperparameters = mean = maximum likelihood estimator of the mean = fluid density, kg=m 3 2 = variancê 2 = maximum likelihood estimator of the variance Subscripts in = diffuser inlet location e = diffuser exit location measured at the engine filter location