I want to build intelligent systems that can communicate with humans effectively and enable individuals to achieve their goals. Today’s systems are often opaque, brittle, and difficult to control, which limits their usefulness in human-centered applications. To make them our trustworthy collaborators, my research aims to (i) understand the computational foundation of generalization in novel scenarios, and (ii) build interactive systems that align with user’s goals.
I am an Assistant Professor of Computer Science and Center for Data Science at New York University. I am affiliated with the CILVR Lab, the Machine Learning for Language Group, and the Alignment Research Group.
Here are some directions I’m excited about nowadays:
- Robustness: Machine learning models are trained on a fixed and often biased dataset, but face a constantly-changing world. How can we build predictors that align with human rationales, avoid spurious correlations, and generalize to out-of-distribution data? How can models adapt quickly given new information?
- Truthfulness: We are increasingly relying on machine learning models (e.g., large language models) for critical tasks. How can we make sure that the model outputs conform to facts? Does the model know what it doesn’t know? Can it output a “proof” for its answer? How do we evaluate factuality efficiently for questions beyond the ability of an average person?
- Human-AI collaboration: We want AI agents to deal with our daily minutiae, support our decision-making, and teach us complex concepts. How should the agent infer user intention and preferences, allow for fine-grained control, and take (natural language) feedback? How will this collaboration shape the future workforce?
Prospective students: Please read my advising statement before contacting. I’m looking for 1–2 PhD students this cycle. If you’re interested, please apply to the PhD program in Computer Science or Data Science and mention my name in your application. If you are interested in a post-doc position, please email me directly. Unfortunately I no longer have time to supervise undergraduate or MS students. If you think you have a compelling story though, feel free to reach out.