Artificial intelligence teaching is being integrated into various disciplines, its theoretical and practical teaching are equally important, and the experimental teaching of artificial intelligence has specific and complex requirements for experimental conditions. This paper introduces the practice and reflection on the method of architecture transformation and construction of AI experimental teaching platform based on traditional laboratory rooms. By fully utilizing existing hardware resources such as workstations for software system architecture implementation and thoroughly introducing open-source technology solutions such as Kubernetes (K8s) and Docker, an all-process AI teaching experiment environment that significantly improves teaching and research efficiency can be created based on the work in this paper. This is achieved without impacting regular teaching activities. The approach optimizes operating mechanisms and service models, realizes demand-customized teaching resources, enhances the level of teaching management, and provides beneficial insights and methods for the construction of an experimental teaching environment for artificial intelligence.