Zailong Wu
Technical Expert at Qingcheng AI
Zailong Wu graduated from the Department of Computer Science at Ocean University of China. His primary research areas include agent prompt engineering and routing optimization, memory semantic–driven distributed computing, many-core architecture optimization, as well as resource scheduling for large-scale clusters in large model training and inference. He has previously worked at the Institute of Information Engineering, Chinese Academy of Sciences, Alibaba Cloud, and China Telecom Cloud, where he has long been engaged in the product development and real-world deployment of artificial intelligence and high-performance computing technologies. He is currently with Qingcheng AI, where he leads the development of the Bagua Furnace intelligent software stack and related agent-based products.
Topic
Agentic AI Infrastructure: An Agent-Centric “Gossip Furnace” Intelligent Software Stack
The integration between the Bagualu Agent and the Bagualu Intelligent Computing Software Stack enables a new generation of agent-centric enterprise software architecture, similar to OpenClaw. The architecture is divided into three layers. Through the Bagualu Agent, the full chain of interaction is realized as: “natural language → Bagualu Agent → Bagualu platform → operation of the Bagualu intelligent software stack → third-party industrial software.” From top to bottom, the three layers of the Bagualu software stack are: The application enablement layer transforms natural language into skill-based capabilities. The core Bagualu engine layer, powered by engines such as Chitu and Mtuner, ensures low-latency inference and real-time agent actions. The compute infrastructure layer (Turnkey) addresses the cost efficiency of running agents.
