Designed for Agent workloads, we are exploring a new paradigm of AI-Native data infrastructure.
As AI evolves from an assistive tool toward autonomous execution, AI Agents impose new requirements on data infrastructure. Unlike traditional human-centric applications, Agents feature frequent small-scale writes, dynamically changing schemas, and hybrid queries over vector and structured data. Traditional evaluation metrics centred purely on throughput and speed are facing new challenges. How to build data infrastructure truly optimised for Agent workloads has become a critical topic in the AI-native era.
Upgrading data infrastructure involves more than database performance. It requires coordination across document parsing, retrieval-augmented generation (RAG), natural language to SQL (NL2SQL), long-term memory, agent runtime and other components. How to shift the focus from raw speed to real-world applicability, and validate technical performance via reproducible benchmarks? Hongqiao International Science and Technology Week 2026 will focus on the transformation of data infrastructure in the AI-native era, discussing full-stack Agent-oriented architectures and their differentiated value in practical deployments.
Scan the QR code to claim your Conference Pass.
Hongqiao International Science and Technology Week 2026 will take place at the National Exhibition and Convention Center (Shanghai) from November 20 to 22. As one of its six thematic stages, the AI Stage gathers industry practitioners across the ecosystem. Through keynote speeches and panel discussions, it examines technical practices, commercial pathways and partnership opportunities during foundation model deployment and industry adoption.
The agenda covers AI technology and product R&D, vertical use-case rollout, enterprise business model transformation, AI investment and financing, as well as AI compliance and governance. It bridges innovation on the technical and industrial sides and delivers multi-dimensional perspectives for industry exchange and business collaboration.
At this forum, Lü Guo, Chief AI Scientist at MatrixOrigin, will deliver the keynote speech Not Faster, but Better Aligned — Paradigm Shift of Data Infrastructure in the AI-Native Era. Addressing new requirements from AI Agents for data read/write operations, queries and runtime environments, he will share insights on the shift of data infrastructure from traditional performance orientation to scenario-adaptive design. He will introduce the full-stack data infrastructure built for Agent workloads, including the MatrixOne database, the MOI Agent data stack, Memoria long-term memory and the Astra Agent runtime. The session will explore how to build Agent-optimised data infrastructure via auditable, reproducible benchmarks and real-world application cases.
Keynote Not Faster, but Better Aligned — Paradigm Shift of Data Infrastructure in the AI-Native Era
As AI evolves from a tool into a creator, the “end users” of data infrastructure are shifting from humans to Agents — and Agents read and write data very differently from people: frequent small writes replace bulk loading, continuous schema evolution replaces static table schemas, hybrid vector-structured queries replace pure SQL, and end-to-end observability replaces post-hoc logging.
This keynote argues that evaluation criteria for data infrastructure in the AI-native era should shift from “faster” to “well-aligned”. Drawing on self-developed benchmark tests (audit-grade reproducible results, public side-by-side comparisons with competing solutions), Lü Guo will present a full-stack data infrastructure tailored for Agent workloads: MatrixOne (Agent-native database), MOI (Agent stack covering document parsing, RAG and NL2SQL), Memoria (long-term memory) and Astra (Agent runtime), together with their unique advantages in live Agent workflows.
The takeaway is not “we are the fastest”, but “we are built for Agent workloads”.
Speech Date: November 20, 2026 Stage: AI Stage
Speaker Bio
Lü Guo MatrixOrigin Chief AI Scientist
Dr. Lü Guo holds a PhD in Pattern Recognition and Intelligent Systems from Shanghai Jiao Tong University, with nearly 20 years of AI expertise — including 8 years of algorithm research and more than 10 years of industry practice. He previously served as AI Solution Expert (Level 19) at Huawei Cloud, Senior AI Product Expert (Level 11) at Tencent Cloud, and Consulting Manager at PwC and IBM. He possesses end-to-end capabilities spanning chips, foundation models, AI platforms and vertical industry solutions. He has led the end-to-end launch of multiple hundred-million-scale AI products, covering manufacturing, semiconductors, finance and government sectors.
Core strengths: He led the planning for Ascend 910/310 foundation model acceleration, delivering over 15% improvement in training and inference performance, and participated in the launch planning of the Pangu multimodal model. He built Tencent Cloud’s TI-ONE / TI-S / TI-Matrix platforms from scratch, serving more than 1,000 enterprise public-cloud customers. He has established AI methodologies for manufacturing, covering quality inspection, anomaly detection, cost forecasting, AI drawing review and FAB movement analysis. Flagship projects include a multi-million-yuan co-innovation AI project with ICBC, Jinpan Technology’s data intelligence and Agent middle platform, production assistance systems for SMIC, and movement prediction & root-cause analysis for Hua Hong. He holds Huawei Trustworthy Certification (Professional Level), Java, PMP and Systems Analyst credentials. He has filed 2 invention patents and published 5 SCI/EI papers.
Register to visit.
Hongqiao International Science and Technology Week ⏰ Nov 20–22, 2026 📍 National Exhibition and Convention Center (Shanghai)
