Beamforming in multi-user scenarios presents unique challenges due to the interference with each other, which can significantly impact signal quality and transmission efficiency. Therefore, in this study, we explore the application of deep learning in designing beamforming for multi-user systems. The study proposes an innovative approach that leverages deep learning to optimize beam patterns for specific applications, control interference, and minimize side effects from unwanted directions. Our approach addresses the challenges of the traditional beamforming method, including adaptive beam pattern planning and real-time adjustment capabilities. The numerical results indicate that our approach outperforms the traditional ones on bit error rate and energy efficiency, and this study also highlights the potential of deep learning to open new pathways for enhancing multi-user systems, aiming to improve transmission efficiency and effectively reduce interference in dynamic environments.