The most time-consuming part of developing new parametrizations for the density functional based tight-binding (DFTB) method consists of producing accurate and transferable repulsive pair potentials. In the conventional approach to repulsive parametrization, every possible diatomic combination of the elements covered by the parametrization must be individually hand-constructed. We present an initial attempt to automate some of this time-consuming process. We consider a simple genetic algorithm-based approach to the fitting problem.