I will start as a Assistant Professor at the Cheriton School of Computer Science at the University of Waterloo and a Faculty Member at Vector Institute in July, 2024. Until then, I am working as a Postdoc Researcher at Microsoft Research New York City.
I am actively looking for strong and motivated students with research experience in Machine Learning and Natural Language Processing.
If you are interested in working with me, please complete this form.
I am on Twitter but email is preferred at
Please apply through the University of Waterloo CS PhD program and mention me in your application.
I cannot take on more interns/external collaborations at this time, but will be able to do so in 2024.
My research is on Reading to Learn: how we can use language understanding to generalize and learn more efficiently. Why should we read to learn? Most machine learning techniques train on vast amount of labeled data or experience for specific problems. When the problems change (e.g. dialogue on a new topic, language interface for a new database, policy for a new environment), we find that the expensive solution we trained no longer generalizes to new problems. Humans generalize to new problems efficiently by reading. For instance, if we buy a new board game, you and I can read the manual and quickly figure out how to play the game, without watching millions of games played by experts. Similarly, we can read the manual of a new coffee machine to figure out how to make coffee with it. The thesis of my research is this:
By reading language specifications that characterize key aspects of the problem, we can efficiently learn solutions that generalize to new problems.
My work in reading to learn spans several areas, including interactive learning, dialogue, question answering, semantic parsing, and knowledge base construction/inference. My recent work has focused on