With the rapid development of artificial intelligence technology, deep learning, as one of its core technologies, occupies an important position in the cultivation of applied talents. Based on the concept of integration of industry and education, this paper proposes a systematic teaching reform plan to address the issues of disconnection between theory and practice, single teaching methods, and insufficient practical resources in the deep learning courses for professional master’s students at our university. Through deep cooperation with Huawei Cloud Technologies Co., Ltd., we introduce cutting-edge theoretical content (such as GoogleNet, ResNet, Transformer, BERT, etc.), update practical cases (covering computer vision, natural language processing, and smart manufacturing), and adopt a case-led comprehensive teaching method combined with the online and offline hybrid practical platform ModelArts to promote the close integration of theory and practice. Simultaneously, a diversified evaluation system with practice as the core is constructed to comprehensively assess students’ practical abilities and project execution levels. The research in this paper provides a valuable reference for the innovation of teaching modes and the cultivation of practical abilities in deep learning courses in higher education institutions.