Diffuse optical tomography (DOT) can be considered as an optimization problem, in which the minimum of an objective function is sought. The objective function is typically some measure of the difference between the predicted and experimentally obtained detector readings. Most of the optimization techniques that are currently applied in optical tomography employ so-called gradient methods. These methods start from an initial guess of the distribution of optical properties and iteratively update this initial guess along the gradient of the objective function. It is well known that the success of gradient techniques depends strongly on the initial guess. If the guess is not chosen appropriately, the algorithm may not converge or may converge to a local minimum. Evolution strategies are global optimization techniques that depend much less on initial guesses. The drawback of evolution-based codes is that they are computationally expensive. In this work we introduce a hybrid approach that combines the advantages of gradient techniques and evolution strategies. The hybrid algorithm is less dependent on an initial guess and overcomes the computational burden connected to evolution strategies.