Abstract: A key component in bringing new piston engine and gas turbine technologies to market is their design optimization. In doing this, engineers need to select the best design parameters out of many alternatives, where each alternative features a unique combination of the design parameters of interest. This talk discusses a novel adaptive surrogate-based scheme known as ActivO, that significantly accelerates design optimization. The proposed approach combines a high-bias learner and a high-variance learner to identify promising regions within the design space and to determine the exact location of the optimum within promising regions, respectively. The results are demonstrated for a range of test surfaces and subsequently, for simulation-driven optimization of an internal combustion engine where the objective is to minimize fuel consumption while adhering to desired emissions and pressure constraints. ActivO is compared to conventional optimization schemes, showing that it reduces the number of simulations needed to reach the global optimum by up to 80%.