Trustworthy AI / Robust AI focuses on building AI systems that operate reliably, fairly, and transparently in real-world environments. This includes ensuring Robustness against noise, distribution shifts, and adversarial attacks, mitigating bias to prevent unfair outcomes across different user groups, and enhancing Explainability so that the reasoning behind model decision is interpretable to humans. By strengthening transparency and stability, these technologies enable users to trust AI systems even under uncertainty. Ultimately, Trustworthy and Robust AI provides the foundational principles needed for the safe, responsible and reliable deployment of AI across high-impact domains.

LLM and Agentic AI explores large language, vision-language, and multimodal foundation models as the core engines for autonomous, reasoning capable AI systems. This area focuses on developing agentic capabilities (e.g., planning, tool-usage, multi-step reasoning, and self-directed decision making) built on top of powerful pre-trained models. By integrating multi-agent cooperation, debate mechanisms, environment interaction, and efficient model adaptation, our research aims to create AI agents that can understand complex contexts, break down tasks, and act intelligently to solve real-world problems. This direction bridges foundational model research with next-generation autonomous AI systems that go beyond passive prediction toward active problem-solving and adaptive intelligence.

AI for X explores how intelligent, domain-adapted AI systems can solve complex, real-world problems, with a primary focus on the energy sector. In particular, this direction develops energy-specialized AI Agents that combine high-trust, physics-based legacy tools with modern AI models to address mission-critical challenges in power system optimization, forecasting, stability analysis, and anomaly detection. By designing agents that can coordinate engineering solvers, simulations, and multimodal data in an integrated workflow, this line of research enables more autonomous, reliable, and efficient energy operations. While centered on energy, these agentic frameworks and methodologies naturally extend to other domains, providing a scalable foundation for AI-driven solutions across diverse industrial and societal applications.