The field of education faces problems such as lack of personalized teaching and insufficient monitoring of students’ learning status. This paper is committed to designing and implementing an intelligent education assistance software that analyzes students’ learning history, ability performance and interest preferences to intelligently recommend personalized learning resources and customized learning paths to optimize students’ learning experience. This paper adopts a model combining DNN and RNN to capture the time series characteristics of students’ learning behavior. In terms of software implementation, a system architecture including front-end user interface and back-end data processing and model inference services is constructed. The front-end adopts responsive design; the back-end is based on microservice architecture. The trained model is deployed on the cloud server and interacts with the front-end through the API interface. At the same time, a user feedback mechanism is set up to regularly collect new data and conduct incremental learning or retraining of the model. After the software is launched, students can log in to the system to see the recommended content customized based on their personal learning profile. After using the software, the students’ maximum average test score increases from 84.9 before use to 99.6, the lowest score increases from 70.3 to 79.9, and the course completion rate increases from 74.925% to 91.45%. These data show that the software can improve students’ learning effects and motivation, and provide support for the intelligent development of educational software.