Aiming at the low degree of industry-education matching, lagging technology iteration, single dimension of practice evaluation and other core problems currently prevailing in network engineering majors of applied undergraduate colleges and universities, we aim to build a four-stage spiral practice teaching system empowered by "industry-university-research-use" in-depth synergy and generative AI technology, and to explore ways to solve the problem of disjointed technology supply in traditional practice teaching. By integrating the resources of the government, universities and enterprises, a four-stage progression path of "basic experiment, comprehensive practical training, project battle and industrial application" is designed, forming a closed loop of ability cultivation. This study proposes a new engineering practice teaching paradigm driven by generative AI, which provides a replicable technical solution and collaborative mechanism for similar specialties.