Master's student in Electronic Information (Artificial Intelligence) at Fudan University, Shanghai.
I am interested in building AI agents that improve reliably through execution feedback. My current focus is on verifiable improvement loops for agent systems: diagnosing failures, proposing bounded changes to tools or harnesses, evaluating them on held-out tasks, and retaining changes only when they pass regression checks.
- AI4AI and self-improving agents
- Agent harnesses: tool use, memory, verification, retry, and rollback
- Code LLM data, evaluation, and agent trajectories
- LLM post-training and agentic reinforcement learning
- Efficient inference and AI systems
- Fudan University — Master's student in Electronic Information (Artificial Intelligence), 2026–present
- Jinan University — B.Eng. in Artificial Intelligence, 2022–2026
- Email: jianyuanlin26@m.fudan.edu.cn