My research focuses on self-evolving language agents, graph world models, and trustworthy LLM systems. I study how agents can learn from experience, external knowledge, and verified feedback to improve their memory, reusable skills, and decision-making — and how world models can support algorithm design on complex networks.
Across these directions I build methods and evaluations for systems that continually adapt to new tasks and environments, with an emphasis on the reliability of both their behaviour and their claimed improvements.
01
AI Applications
Putting agents to work in domains where being wrong is expensive.
- Finance AI
- Education AI
02
AI Foundations
The machinery underneath: verification, self-evolution, world models, and infrastructure.
- Reliability & Verification
- Self-Evolving Agents
- World Models for Networks
- Efficient Agent Infrastructure