This second paper presents the aerothermal optimization of the first stage rotor blade of an axial HP turbine by means of the conjugate heat transfer method (CHT) and lifetime model described in paper 1. The optimization system defines the position and diameter of the cooling channels leading to the maximum lifetime of the blade while limiting the amount of cooling flow. It is driven by the results of a CHT and subsequent stress analysis of each newly designed geometry. Both temperature and stress distributions are the input for the Larson-Miller material model to predict the lifetime of the blade. The optimization procedure makes use of a Genetic Algorithm (GA) and requires the aerothermal analysis of a large number of geometries. Because of the large computational cost of each CHT analysis, this results in a prohibitive computational effort. The latter has been remediated by using a more elaborate optimization system, in which a large part of the CHT analyzes are replaced by approximated predictions by means of a meta-model. Two metamodels, an Artificial Neural Network (ANN) and a Radial Basis Function (RBF) network, have been tested and their merits have been discussed. It is shown how this optimization procedure based on CHT calculations, a GA and a metamodel can lead to a considerable extension of the blade lifetime without increase of the amount of cooling flow or the complexity of the cooling geometry.