I am a co-founder of and Principal Research Scientist at Bioptimus, working on building the first universal foundational model for biomedical applications. Prior to this, I was a Senior research scientist at Google DeepMind, where I studied how the diversity of a model’s predictions informs the models’ own uncertainty. I received my PhD from MIT, working with Suvrit Sra and as a member of the Machine Learning and Learning and Intelligent Systems groups. My PhD work focused on deriving and sampling from negatively dependent measures, which use Strongly Rayleigh polynomials to encode desirable properties in distributions over subsets of diverse items.

MITの大学院生の時に、Suvrit Sraと一緒に研究していた。機械学習の数学的な分析と非凸最適化を研究している。その他にも、理論計算機科学や位相幾何学や言語学に興味がある。

Recent(ish) news

Projects

ML for biology

ML for biology

Using ML to design novel proteins

Image credit: Shutter2U/Getty Images

Modeling uncertainty

Modeling uncertainty

Asking a machine learning model to quantify its uncertainty about an output.

Plot adapted from the sklearn GP Regressor example.

Strongly Rayleigh measures

Strongly Rayleigh measures

A class of probability measures over subsets of a ground set that enable negative dependence: similar items are pushed away from each other.

Image courtesy of NASA