Under the BD-driven research on multimodal data analysis and intelligent system design, the work focuses on multimodal data analysis and intelligent system design, with BD-driven as the core. The system consumes a variety of data types from different sources, such as text, images and audio, to construct a large-scale and diverse data set. Then, the system cleans the data and pre-process the data to ensure that the data is accurate and consistent. Bluemix can be used to organize the services of different modal data, although it is easy to expand. After multimodal data fusion, it can also organically combine the features of different modal data and get a more comprehensive and abundant data representation form. Experiments show that the processing accuracy, data integration rate, and data utilization rate of the BD-driven system for modal data analysis and intelligent system design are all above 89.27% and the manual cooperation rate is 11.02%, which is the best of all the BD-driven and non-BD-supported intelligent system driven systems. It can be concluded that the system model has passed the evaluation and optimization, and is successfully applied to practical problems, and provides strong support for intelligent decision-making.