AI Agents for Computer Systems Research

Collaborative AI agents plan, implement, and verify research software. Code, experiments, and results go to human review; questions and corrections feed back into the next cycle.

Research questions

  • How should agent teams divide the work of developing and maintaining research software?
  • When should an agent ask a human to check a result or make a decision?
  • What evidence do humans need to verify an agent-run experiment?
  • How can humans and agents explore architecture designs together without losing track of why a design was chosen?

Publications

2026

ASE 2026
TheBotCompany: Self-Organizing Multi-Agent Systems for Continuous Software Development
Wenhan Lyu, Yue Xiao, Yixuan Zhang, Yifan Sun

2025

CAMS 2025
DaisenBot: Human-AI Collaboration in GPU Performance Analysis with Multi-Modal AI Assistant
Enze Xu, Jeremy Coonley, Daoxuan Xu, Yifan Sun