UPDATE: I am working on something new related to data and machine learning. If you are an experienced research scientist and/or engineer who wants to build tools to improve deep learning models by intervening on training data, please get in touch.
My primary research interest is furthering the science of deep learning—developing and conducting rigorous, empirical experiments to better understand deep learning systems—and using the resulting insights to improve the performance, efficiency, and human-interpretability of these systems. I used to pursue these interests as the Head of Data Research at MosaicML.
I also spent 1.5 years as an AI Resident at FAIR, where I applied empirical approaches to understand deep learning systems under the supervision of Ari Morcos and Sergey Edunov. The publications section of my site has summaries of all my work.
For my doctorate I studied the neurobiology of cognition at McGill University in Montréal with Julio Martinez-Trujillo. A friend once described my work as “translating weak electrical discharges in the brain into strong scientific statements about the mind”.
Both my neuroscience and machine learning research have been motivated by an interest in how complex systems and high-dimensional information can be represented humanely and communicated meaningfully.
I’m a contributor to vissl, FAIR’s library for self-supervised learning.