Shaonan Ma
Head of the Data Intelligence Team at Qiyuan Lab.
He holds a Ph.D. from the Institute of High Performance Computing at the Department of Computer Science, Tsinghua University. He has published dozens of papers in leading international conferences in computer science, including FAST, SIGMOD, SC, ASPLOS, ICDE, ICPP, and Cluster, as well as in international journals such as TPDS and TKDD. He has led one key project under a major national research program. He led the development of the Jiuyuan Intelligent Graph Database for Inference (JYGraph). Built on the domestic Phytium platform, the system achieved top-tier performance in the internationally recognized LDBC benchmark. He also led the development of the Jiuyuan Intelligent Big Data Machine system, which adopts a “storage-for-compute” paradigm to reduce computing costs while improving inference performance. The system has significantly outperformed existing solutions across multiple academic benchmarks and real-world business scenarios.
Topic
Hardware–Software Co-Designed Computing Architectures and Intelligent Data Integration Practices for the Era of Large Models
Building autonomous and controllable foundational infrastructure is critical for AI development. This presentation will focus on Qiyuan Lab’s **Jiuyuan Intelligent Software Stack**, sharing how we are establishing a domestically developed intelligent computing foundation. The content covers two main aspects: 1. **Jiuyuan Intelligent Computing Platform** – By designing a unified acceleration platform on heterogeneous domestic computing resources, it abstracts away chip differences and provides stable and efficient training and inference support for large models. 2. **Jiuyuan Intelligent Data Integration Platform** – Leveraging hardware–software co-design, it connects the full technical chain from low-level data processing and knowledge base construction to intelligent analytics. Through breakthroughs in underlying architecture, Jiuyuan is building a new ecosystem for intelligent computing, supporting enterprises in their AI-driven transformation. **Outline:** * Computing and data challenges in the era of large models * Computing foundation: Jiuyuan Intelligent Computing Platform * Data foundation: Jiuyuan Intelligent Data Integration Platform * Ecosystem and outlook: building the future of domestically developed intelligent computing **Audience Takeaways:** * Understand how system-level design can address ecosystem fragmentation and model migration challenges posed by heterogeneous domestic computing resources. * Gain insights and practical experience in building an integrated intelligent data system on domestic computing and storage hardware.
