I build AI agents that run industrial workflows, and dexterous robot hands that give embodied AI the ability to work with its hands.
Visit WillFlow.aiPoint at the palm to grasp. Hold to grip.
I am a Research Scientist at Shanghai Jiao Tong University and the founder of WillFlow.ai. My work sits where AI meets physical industry: agents that plan and carry out real tasks in factories, and dexterous hands that let robots manipulate the world the way people do.
LLM-based agents that read production and equipment data, plan maintenance and operations tasks, and act through the software a plant already runs, with the reliability and safety checks industry demands. This is what we build at WillFlow.ai.
Multi-fingered robot hands that grasp and manipulate objects in-hand: learning control policies, using touch as feedback, and bringing skills learned in simulation onto real hardware.
Earlier work. During my Ph.D. at UMass Lowell with Dr. Yimin Chen, I studied the security of machine learning: backdoor attacks and defenses in federated learning, robust model merging, and privacy risks in RAG and LLM agents. Before that, at SUSTech with Dr. Yinqian Zhang, I worked on blockchain security, including Ethereum private transactions and smart contract vulnerabilities.
Buffer is All You Need: Defending Federated Learning against Backdoor Attacks under Non-IIDs via Buffering
Xingyu Lyu, Ning Wang, Yang Xiao, Shixiong Li, Tao Li, Danjue Chen, Yimin Chen
BoBa: Boosting Backdoor Detection through Data Distribution Inference in Federated Learning
Zhengyuan Jiang*, Xingyu Lyu*, Shanghao Shi, Yang Xiao, Yimin Chen, Thomas Hou, Wenjing Lou, Ning Wang
Beyond Uniformity: Robust Backdoor Attacks on Deep Neural Networks with Trigger Selection
Shixiong Li, Xingyu Lyu, Ning Wang, Tao Li, Danjue Chen, Yimin Chen
Mengya Zhang, Xingyu Lyu, Jianyu Niu, Xiaokuan Zhang, Yinqian Zhang, Zhiqiang Lin
Adversarial Attacks on Federated Learning Revisited: a Client-Selection Perspective
Xingyu Lyu, Shixiong Li, Ning Wang, Tao Li, Danjue Chen, Yimin Chen
An Empirical Study on Ethereum Private Transactions and the Security Implications
Xingyu Lyu*, Mengya Zhang*, Xiaokuan Zhang, Jianyu Niu, Yinqian Zhang, Zhiqiang Lin
* Equal contribution
Research Scientist, Shanghai Jiao Tong University
Embodied AI with dexterous robot hands, and AI agents for industry. Shanghai.
Founder, WillFlow.ai
Building AI agents for industrial workflows. Shanghai.
Visiting Student, Zhejiang University
Robust model merging. Hangzhou, Sep to Dec 2025.
Research Intern, Dewave Ventures
Real-world assets (RWA) and blockchain. Shanghai, Sep to Dec 2025.
Research Assistant, University of Massachusetts Lowell
Security and privacy in AI with Prof. Yimin Chen: attacks and defenses in federated learning, RAG and LLMs. Lowell, MA.
Research Engineer, Heurist.AI
DeFi and MEV security research. New York, Jun to Aug 2024.
Visiting Student, HKUST (Guangzhou)
LLM security with Prof. Xinlei He: privacy attacks on RAG and mitigation in agent systems. May to Jun 2024.
Research Assistant, Southern University of Science and Technology
Blockchain security with Prof. Yinqian Zhang: smart contract vulnerability detection, DeFi and MEV. Shenzhen.
Visiting Student, City University of Hong Kong
Adversarial attacks in autonomous vehicles with Prof. Jianping Wang: sensor attack detection and resilience.
Visiting Student, Pengcheng Laboratory
Cyber range project: blockchain security testing and simulation. Shenzhen.
Detection method, terminal and storage medium for Ethereum private transactions
CN115187250B, China. Yinqian Zhang, Xingyu Lyu, Jianyu Niu
Real-time detection method, terminal and storage medium for Ethereum smart contract vulnerabilities
CN115174279B, China. Yinqian Zhang, Xingyu Lyu, Jianyu Niu
UMass Professional Development Fund, Graduate Research Grant
UML Graduate Research Grant Award
Moloney Scholarship
ACSAC Student Travel Grant
Ph.D. in Computer Science
University of Massachusetts Lowell, USA
M.Sc. in Computer Science
Guangzhou University, China
B.Sc. in Electrical Engineering
Shanghai University, China