🎓 About Me
I am a fourth-year undergraduate student at 
Tongji University. During my undergraduate studies, I spent one semester as an exchange student at 
the University of California, Berkeley, and participated in the AI Summer School at 
Westlake University.
Incoming graduate student with Master of Science in Computer Science in UC, San Diego 
(27Fall)
My research interests are AI Agent, Agent Safety, Embodied AI, AI for Software Engineering (AI4SE), and Human-Computer Interaction. I was a research assistant at the SoftACE Lab, Tongji University, advised by Prof. Yan Liu. Moreover, I am also interested in the intersection of AI and transportation, as I previously studied Civil Engineering for one year at Tongji University.
🔥 News
- 2025.06: 🔥🔥 My first co-authored paper with Peiran Li under supervision of Prof. Zhengzhong Tu is on Arxiv, SAFEFLOW: A Principled Protocol for Trustworthy and Transactional Autonomous Agent Systems
- 2025.05: 🔥🔥 Successfully defended undergraduate thesis titled “Research and Application of a Multi-Agent-Based Agile Requirements Elicitation Methodology”(基于多智能体的敏捷需求诱导方法研究与应用) with a perfect score (5/5)
- 2025.03: 🔥🔥 My first paper as a first author is on Arxiv, Goal2Story: A Multi-Agent Fleet based on Privately Enabled sLLMs for Impacting Mapping on Requirements Elicitation.
- 2024.12: 🎉🎉 Happy to join the AGI Lab, Westlake University as a visiting student, under supervision of Prof.Chi Zhang.
- 2024.05: 🎉🎉 Happy to join the Future Lab, Tsinghua University as a research assistant, collaborating with Jiachen Du.
- 2023.12: 🎉🎉 Happy to join the Tongji-MIT City Science Lab as a research assistant and developer, collaborating with Chance Jiajie Li.
📝 Publications
SAFEFLOW: A Principled Protocol for Trustworthy and Transactional Autonomous Agent Systems
Peiran Li*, Xinkai Zou*, Zhuohang Wu, Ruifeng Li, Shuo Xing, Hanwen Zheng, Zhikai Hu, Yuping Wang, Haoxi Li, Qin Yuan, Yingmo Zhang, Zhengzhong Tu
*Equal contribution
- SafeFLow: principled protocol-level framework explicitly designed to build secure, reliable, and trustworthy LLM/VLM-based autonomous agents
- SafeFlowBench: the first-of-its-kind benchmark specifically designed to evaluate VLM-based agent safety in GUI-based environments
Xinkai Zou, Yan Liu, Xiongbo Shi, Chen Yang
- Goal2Story: first work to realize goal-driven requirements elicitation via LLM-based agents
- StorySeek: first dataset constructed specifically for goal-driven requirements elicitation with other potential information
🎖 Honors and Awards
- 2023 Tongji University International and Hong Kong-Macao-Taiwan Exchange Scholarship
- 2021 Tongji University Undergraduate Outstanding Student Scholarship
- 2021 Tongji University Student of the Year
📖 Educations
- 2021.09 - 2025.07, B.Eng. in Software Engineering, Tongji University
- 2023.08 - 2023.12, Exchange Student, University of California, Berkeley
🧩 Projects
A project aiming at encouraging greener daily commute behaviors. SoCity Green Commute explores personal carbon allowance, tokenomics, and decentralized governance using a mobile DApp.
Collaboration with Chance Jiajie Li, Ruyi Yang, Zhenze Mo, etc.
ArchInspire (under revision)
This paper introduces an AI-augmented design copilot that enhances conceptual abstraction by leveraging extensive datasets, advanced annotation using GPT-4V, and custom generative models, all seamlessly integrated to provide three key functions: an architectural search engine, bidirectional object-based architectural analysis, and an architectural design generator, creating a powerful tool for architectural inspiration.
Collaboration with Ruyi Yang.
BCG Gen-AI Platform
This project is a platform for BCG employees, consultants and clients to use privately deployed LLMs and tools.
💻 Internships
- 2023.12 - 2024.05, Data Engineer Intern (Under supervision of Tianyu Shi, world’s no.1 mentor),
Tesla (Shanghai Gigafactory), China. - 2023.07 - 2023.09, Software Development Engineer Intern,
Zhejiang Smart Port Technology Co., Ltd., China.