I am a PhD Candidate in Computer Science at Columbia University advised by Prof. Smaranda Muresan. My research advances sociotechnical AI alignment: measuring and optimizing the societal impact of AI to build models that preserve the diversity of human values and ideas as well as people’s ability to understand and steer the systems they rely on. Combining deep learning methods with insights from disciplines such as linguistics, psychology, sociology, and political science, I have approached these challenges along three directions:
- Pluralistic & Global Value Alignment: Quantifying how well models align with their sociocultural context [1, 2, 3, 4].
- Pragmatic Understanding & Expert-AI Collaboration: Developing human-AI collaboration frameworks, such as in-context learning with expert feedback [5], to measure and enable understanding of implicit knowledge required to interpret scientific entailment [6], underspecification [7], stylistic appropriateness, sociocultural norms [2, 3, 4], and figurative meaning [8, 9, 10, 11], across text, images, and videos.
- Meta-evaluation & Metric Design: Developing theoretically grounded and sound evaluation frameworks for complex human-centered constructs such as creativity [12].
During my graduate studies, I interned at Google DeepMind and at AWS AI. Before starting my PhD, I completed my B.A. at Columbia College and worked as a data scientist at UBS.
