With the rapid advancement of artificial intelligence (AI), traffic engineering is undergoing a critical transformation that requires restructuring both its knowledge framework and talent cultivation model. Traditional railway-related curricula struggle to address highly complex and real-time problems such as transportation organization, train control, and capacity prediction, resulting in fragmented competencies, outdated modules, and misalignment between instructional content and operational needs. As a data-intensive and safety-critical subsystem, modern railway operations increasingly rely on AI for train control, dispatch scheduling, operational optimization, and digital infrastructure management, calling for systematic curriculum reform. In response, this study examines global teaching practices and technological trends and proposes the AITL Layered Curriculum Model, a three-stage competency pathway encompassing technology internalization, scenario transfer, and intelligent creation. The model establishes an integrated instructional content chain, a task-driven mechanism, and an AI-enabled experimental platform embedded in representative railway scenarios. Multi-scenario virtual teaching experiments verify that the AITL model effectively mitigates fragmented curriculum organization, insufficient task embedding, and unclear competency progression, providing a systematic, transferable, and evaluable framework that aligns with the intelligence-oriented transformation of railway traffic engineering and offers broader applicability to transportation education.