/PRZWT/At the "China AIGC Innovation Application Forum" at the China International Trade in Services Fair in 2025, Gao Xuefeng, founder and CEO of Fengqing Technology, systematically explained the technical logic and industrial practice of "data-centric construction of AI agent platform" to global participants, revealing a new path for artificial intelligence empowering the real e****my. The company has served more than 30 state-owned central enterprises and industrial leaders in the fields of chemical industry, manufacturing, energy, finance, etc., and built more than 30 high-value application scenarios. Its "cloud-side-end" collaborative architecture is becoming a key support for industrial intelligence upgrading.
Data as the core: cracking the big model landing dilemma
Gao Xuefeng pointed out that the current generative AI faces four core challenges: model illusion, weak reasoning ability, poor interpretability, and data security risks. Fengqing Technology proposes a "data-centric" technology path, which converts enterprise high-value data into callable expertise through two-wheel drive of knowledge engine and industry distillation model. Its "Fengqing · Enterprise Knowledge Middle Platform" supports the efficient integration of structured, semi-structured, and unstructured data, and builds an enterprise-specific knowledge base through graph and vector fusion technology to achieve precise scene-based applications.
Take the group's artificial intelligence platform built in cooperation with Sinochem in China as an example. The platform realizes the intelligent transformation of chemical and agricultural scenes through the cloud-side collaborative architecture. The cloud deploys a multi-modal knowledge base, the side carries departmental-level shared nodes, and the end side is equipped with personal exclusive agents to form an intelligent system of "knowledge on-demand switching and computing power on-demand call". Practice shows that the platform improves the accuracy of document recognition to 99.2%, the efficiency of equipment predictive maintenance is increased by 40%, and the response time of the security operation manual SOP intelligent question answering system is shortened to 0.3 seconds.
Scenario landing: from leading enterprises to small and medium-sized enterprises
Fengqing Technology realizes technology inclusion through a "three-step" strategy: the first step is to build high-value scenarios with industry leaders, such as cooperating with China Resources Pharmaceutical to achieve intelligent questions and intelligent indicator analysis in the field of biomedicine; the second step is to push mature scenarios down to end point agents through cloud replication, such as the AI empowered new material research and development joint laboratory jointly established with Sinochem Information and Jilin University at the beginning of the year, whose R & D agents have been applied to 37 new material research and development projects; the third step is to achieve C-end inclusive through personal exclusive agents, and its Fabarta products launched in July 2025 have covered writing and translation
In high-frequency office scenarios such as data analytics, the end side preferentially calls the local model, and requests the cloud side when computing power is required, which not only guarantees privacy but also improves efficiency.
Ge Shuang, co-founder and COO of Fengqing Technology, added: "Large central state-owned enterprises hope to achieve multi-scenario intelligence through the platform, such as production line workers performing predictive maintenance of equipment through personal maintenance super agents. Small and medium-sized enterprises can quickly replicate the capabilities of leading enterprises through lightweight agents. The three-level'cloud-side-end 'combined services we launched have been adapted to the needs of enterprises of different sizes." Data show that this strategy has enabled Fengqing Technology to triple its operating income this year.
Industry Enlightenment: A New Paradigm for Intelligent Upgrading
In terms of global layout, Fengqing Technology has built a cross-border agent matrix through its holding subsidiary Fengqinghaitu, covering core links such as customs declaration and tax refund, compliance risk control and so on. Its agent acceleration platform jointly built with Volcano Engine has supported multi-language cross-border business processing, and the efficiency of customs declaration processing has increased by 60%. Gao Xuefeng said: "We adhere to the technical philosophy of'cloud-side collaboration ', integrate knowledge through the cloud, empower organizations at the edge, and liberate individuals at the end point, and finally realize industrial upgrading."
Facing the future, Fengqing Technology will deepen the "artificial intelligence +" action strategy and continue to make breakthroughs in the field of AI + science and technology. The agent acceleration platform it cooperates with China Electric Power Science and Technology Institute has been applied to scientific research institutions to accelerate the transformation of innovative achievements; in the field of people's livelihood and well-being, user feedback from legal, educational, scientific research institutions and other fields shows that personal exclusive agents save 30% of white-collar office hours and improve efficiency significantly.
From data association analysis in the traditional BI era, to the technical explosion of computer vision, and then to the industrial empowerment of generative AI, the development of artificial intelligence has always followed the evolution logic of "technology-scene-ecology". The practice of Fengqing Technology reveals that the data-centric technology path can effectively solve the problems of model illusion and data security; the cloud-side collaborative architecture can empower the whole chain from leading enterprises to small and medium-sized enterprises; and the C-end universal benefit of personal exclusive agents builds a virtuous circle of "experience-transformation-growth".
As Gao Xuefeng said at the forum: "Artificial intelligence is not to replace people, but to liberate people. By building a'data-centric 'agent platform, we are enabling every individual, every organization, and every industry to have an exclusive intelligent partner." This new paradigm of industrial intelligence with data as the core, scene as the wing, and ecology as the vein may become a key engine for the high-quality development of China's real e****my in the next decade.
