Zeyi Chen profile photo

Zeyi Chen

陈泽奕

ECE Student

Shanghai Jiao Tong University

UM-SJTU Joint Institute

CMU Exchange, Spring 2026

I am an undergraduate student in Electrical and Computer Engineering at the UM-SJTU Joint Institute, Shanghai Jiao Tong University. I was an exchange student at Carnegie Mellon University in Spring 2026. My research interests lie in building persistent, adaptive AI systems that coordinate with humans and other agents over long time horizons.

  AI Agents and Agentic Web Infrastructure
  1. Build web-scale infrastructure for persistent LLM-based agents that can remember, use tools, and collaborate across long-running tasks.
  2. Study how agent systems should expose skills, memory, permissions, and coordination protocols to humans.
  3. Connect research prototypes with real systems through Holos and agentic web infrastructure.
  Multi-Agent Systems and Agent Societies
  1. Model how local interaction rules produce emergent collective behavior in groups of adaptive agents.
  2. Study communication, coordination, cooperation, and belief formation in agent societies.
  3. Use simulation to understand how protocols and incentives shape system-level dynamics.
  Reinforcement Learning for Emergent Behavior
  1. Apply reinforcement learning to mixed human-autonomous systems such as traffic networks.
  2. Analyze how learned policies interact with human-like behavior and affect global outcomes.
  3. Design adaptive systems that remain interpretable, steerable, and robust in social environments.

News

Selected Research Projects

Holos / Agentic Web Infrastructure

Persistent LLM-based agents, tool use, memory, permissions, and human-agent collaboration.

Research project · Shanghai Innovation Institute

Mixed Human-Autonomous Traffic Systems

Reinforcement learning and simulation for emergent driving behavior in traffic systems with both human and autonomous vehicles.

Manuscript under review

Opinion Dynamics in Agent Societies

Modeling how interaction structures and adaptive agents influence belief formation and social dynamics.

Research direction

Get Connected

Fellow researchers: I am happy to connect over AI agents, multi-agent systems, reinforcement learning, social simulation, and long-term human-AI collaboration.

Students and builders: If you are working on agent infrastructure, adaptive systems, or research tooling, feel free to reach out.