CV

I am a 3rd-year Ph.D. student in Computer Science at the University of Maryland (UMD), advised by Prof. Soheil Feizi. My research focuses on understanding, evaluating, and improving the reliability of LLMs and LLM agents across four connected areas:

• Reliable Agents & Tool Use: Understanding how agents select tools, decide when tools are needed, and translate those decisions into reliable actions.
• LM Interpretability & Control: Explaining model behavior through internal representations and reasoning processes, and understanding when and how interventions can steer it.
• Red Teaming & Reliable Evaluation: Exposing failures through adversarial testing and contamination detection, and measuring capability breadth and consistency beyond aggregate accuracy.
• Reasoning: Studying how models organize reasoning, allocate computation across questions, and explore interactive environments.