Beijing Zhongguancun Kejin Technology Co., Ltd. is a leading company specializing in large language model (LLM) technologies and applications. Leveraging three core technologies—proprietary domain-specific large language models, big data analytics, and multimodal interaction—the company has developed a range of product solutions, including the Dezhu Large Model Platform, Intelligent Marketing, Intelligent Customer Service, Intelligent Operations, and Knowledge Management. It has also established its next-generation intelligent interaction brand, “Dezhu Intelligence,” helping enterprise clients drive innovative AI applications.

I. Challenges Faced by Ningxia Jiaojian Transportation Technology Research Institute in Building an LLM Application Development Platform

As a leading transportation technology research institute in the region, Ningxia Jiaojian Transportation Technology Research Institute has increasingly recognized the significant potential and value of large language model (LLM) technology in areas such as architectural design, construction management, and infrastructure maintenance as its digital transformation deepens. However, in practical applications, Ningxia Jiaojian faces several challenges and pain points: general-purpose models cannot adequately support complex, highly specialized scenarios; there is a shortage of technical talent and domain-specific capabilities; internal knowledge is difficult to search and manage; and preparing tender documents is time-consuming and prone to errors.

II. Solutions Provided by Zhongguancun Kejin

• Unified management of customers’ private models and various open-source models, enabling domain-specific training and application development.

• Rapid development of LLM applications such as AI+BI, knowledge assistants, and tender assistants, enabling second-level access to knowledge and improving efficiency in bidding and tendering scenarios.

• Hands-on support that helps customers’ technical teams build a closed-loop capability across LLM application evaluation, development, and optimization.

III. Results

• Significantly improved the efficiency of scenario validation, increasing LLM application development efficiency by more than 50%.

• Improved enterprise knowledge search efficiency by 50%, while reducing the operational burden of the knowledge base by 70%.

• Reduced tender document generation time by more than 70%.