Abstract: The rapid development of generative artificial intelligence (AI) is profoundly driving the paradigm shift in higher education. Based on the national first-class undergraduate course “Engineering Mechanics” (in English) at Beijing Institute of Technology, this paper constructs a course-specific AI empowerment path with a three-tier architecture of “knowledge base-enhanced model-AI application" centered around the trinity teaching concept of “AI empowerment + practice orientation + bilingual adaptation". By building a multi-level bilingual knowledge graph in Chinese and English, developing a 24-hour bilingual intelligent learning assistant and exclusive intelligent agents, creating digital human lecturers, and guiding students to independently generate HTML animations through projects, the deep integration of AI tools with the entire teaching process from “before class-during class-after class" is achieved. On this basis, rigorous AI-assisted examination comparison experiments were designed and implemented to quantitatively evaluate the teaching effectiveness of human-machine collaboration. Two independent experiments showed that effective human-machine collaboration can significantly improve exam scores (with a more than 20% improvement in the same group), and the average scores of the current AI-assisted group surpassed the level of previous closed-book exam students. The work presented in this paper provides a practical case for promoting the transformation of fundamental mechanics courses from "knowledge impartation" to "ability construction", and also offers quantitative evidence and empirical references for the construction of smart courses and teaching evaluation reform in the context of new engineering education.