Understanding AI agents to safely utilize them.
AI agents are becoming more and more advanced. To safely utilize them in real-world workflows, we need to understand how they reason, where they fail, and what they are actually doing inside. Our research investigates the internal mechanisms of large AI models — building the foundation for trustworthy, transparent deployment.
Integrating cutting-edge AI agents into your personal workflows.
We are witnessing an age of super platforms proposing increasingly capable AI tools. Integrating these tools meaningfully into individual workflows can significantly enhance user's productivities. We investigate how AI agents can be architected as persistent companions, adapting continuously to each user's context and preferences.