Ph.D. Student · Emory University
LLM Agents · Alignment · AI Safety
I am a Ph.D. student in Computer Science at Emory University. My goal is to turn advances in AI into greater human agency: helping people accomplish complex goals while retaining meaningful control. My research centers on LLM agents, alignment, and AI interpretability and safety, with a particular interest in multi-agent collaboration and theory of mind. I explore how agents can understand their environments, reason about their collaborators’ beliefs and intentions, and work together effectively. I am especially interested in agents that adapt to people’s values and preferences and recognize when to act autonomously or seek human input. Alongside these capabilities, I study how to evaluate and explain agent behavior, and uncover risks that emerge through interactions among agents.
Before my Ph.D., I spent two years building and deploying LLM and RAG systems in industry, including AI copilots and decision-support systems. I hold an M.S. in Data Science from Columbia University and a B.A. in Computer Science & Economics from Emory University.
Always happy to talk about LLM agents, alignment, or AI safety — reach out any time.
How can agents understand their environments and collaborators well enough to plan and work together reliably?
How should agents adapt to individual goals and preferences, and decide when human involvement is most valuable?
How can we explain and safeguard agent behavior, including risks that emerge through shared memory and communication?
arXiv preprint, 2026 (under review)
Helps coding agents find relevant code by bringing repository structure directly into their existing search workflow, supporting codebase understanding and test generation.
IEEE Engineering in Medicine & Biology Conference (EMBC), 2026Oral
Uses fetal and maternal ECG to reconstruct Doppler waveforms and study which aspects of fetal cardiovascular function are recoverable across modalities.
Under review, 2026
Reviews 22 studies of multimodal pretraining for wearable sensor foundation models, organizing five learning paradigms and identifying challenges in personalization, robustness, and generalization.
