This paper presents a optimization method based on genetic algorithms (GA) and deep neural network (DNN) for solving the aerodynamic shape and trajectory co-optimization problem of re-entry vehicles to maximize range. The Gaussian pseudospectral method is first employed to optimize trajectories under different shape parameters, constructing a dataset that maps shape parameters to maximum range. A DNN model is then developed to replace the computationally expensive pseudospectral solving process and is embedded into the GA framework for global optimization. Numerical simulations demonstrate that the DNN model achieves an average absolute relative error below 1%. More-over, it improves computational efficiency by four orders of magnitude compared to the Gaussian pseudospectral method, ensuring both accuracy and efficiency in GA-based optimization. Numerical results confirm that the optimized shape parameters increase the maximum range by 67% compared to the baseline, demonstrating the effective-ness of the proposed method.