The rapid growth of online tutoring systems and computer-aided education, driven by advancements in information technology, has resulted in a wealth of student learning data. This motivates the question on how to achieve personalized learning through an automatic manner. Thanks to the impressive advancement of machine learning in recent years, such as deep learning and large language models, using machine learning techniques for realizing personalized learning shows its promise. However, a prevalent trade-off in current machine learning methods is the compromise between improved performance and interpretability. This is primarily due to the opaque nature of neural networks. But as a high-stake decision-making domain, education requires the methods to have interpretability to be deployed in practical systems. In this talk, I will introduce our technical attempts on how to achieve interpretable machine learning for personalized learning while retaining state-of-the-art performance. Besides, I will show our practical experiences on implementing real personalized learning systems.