With the exponential growth of cross-modal information in railway data-to-service platforms, including accumulated text, images, files, and audiovisual materials, efficiently managing and utilizing these valuable resources has become an important issue for optimizing data organization and improving retrieval efficiency. This paper innovatively proposes a design scheme for an intelligent cross-modal data retrieval system tailored to the railway domain. The system aims to overcome the limitations of traditional single-modal retrieval by integrating multiple data types such as images and files, using names as core indices, to achieve seamless retrieval and integrated presentation of cross-modal information. This system not only accurately retrieves all modal information related to specific individuals, such as academic papers, on-site photos, research reports, and related news links, but also deeply extracts and categorizes key information, such as providing file paths, report titles, and enabling download operations, greatly enhancing the efficiency and precision of railway data utilization and retrieval.