Abstract: Technology transfer is an important link between scientific innovation and industrial innovation, and also a key pathway for promoting the deep integration of the science-industry chain. At present, the scale of technology transfer and the level of patent commercialization in Chinese universities and research institutes continue to improve. However, the connection between technology transfer and industrial demand still faces problems such as inefficient matchmaking, difficulty in discovering suitable achievements, inaccurate matching, and low service efficiency. In the technology transfer practices of new-type research institutions, traditional workflows are still relatively dependent on manual experience, making it difficult for enterprises to identify suitable scientific and technological achievements and research and development teams. In addition, the cost of demand communication and interpretation remains high, and expert selection and resource allocation lack effective intelligent technical support. To address these problems, this paper focuses on the technology transfer practices of the Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, and develops the key technologies of the “Tianji Xing” system, including large language models, knowledge graphs, and talent profiling. The system supports key business processes such as multi-source scientific and technological resource organization, enterprise demand understanding, achievement retrieval and recommendation, and expert team matching. The system has been deployed in technology transfer practice. As of December 2024, the related model has served 1253 enterprises, facilitated the transfer and commercialization of 108 patents from new-type research institutions, and provided application validation in achievement retrieval, expert matching, and technology transfer service workflows.