Abstract: With the rapid development of optical fiber communication technology, improving the coupling efficiency of semiconductor laser and optical fiber has become a key technical issue to enhance the signal transmission capability. Although the traditional simulation optimization method can improve the coupling efficiency, the process is time-consuming and requires rich theoretical and practical experience. In order to solve this problem, this study combined the simulation data and actual parameter data of Zemax optical design software, used the neural network model to fit the mapping relationship between parameters and coupling efficiency. Genetic algorithm was used to optimize the nonlinear mapping obtained by neural network, so as to quickly determine the maximum coupling efficiency and the corresponding parameter settings, avoiding the complicated parameter adjustment process in the traditional method. The simulation results show that the coupling efficiency optimization utilizing the artificial intelligence algorithm can improve the coupling efficiency by 5% compared with the traditional optimization method, while significantly reducing the number of experiments and the cost. This method provides a new idea and tool for the coupling design of semiconductor laser and fiber, and is expected to significantly improve the coupling efficiency and promote the development and application of fiber communication technology.