About Me

I currently lead the CTO research and innovation group at Mobileye under Shai Shalev-Shwartz. Before that, I did my Computer Science PhD at Harvard University where I was advised by Boaz Barak.

My research focus was gaining a better empirical and theoretical understanding of Deep Learning: Searching for the right abstractions to reason about DNNs, what makes them so successful in practice and how can we make models robust and easier to train. Together with some fellow Machine Learning enthusiasts we founded the Harvard Machine Learning Foundations group.

Before the PhD, I worked as Algorithm Developer in Mobileye from 2017-2018. Even before that, I did my undergrad in the Hebrew University of Jerusalem in Mathematics and Computer Science, graduating as valedictorian in 2017.

Email: galkaplun (at) g.harvard.edu

Google Scholar

About Me

Publications

Deconstructing Distributions: A Pointwise Framework of Learning

Knowledge Distillation: Bad Models Can Be Good Role Models

For self-supervised learning, Rationality implies generalization, provably

For manifold learning, deep neural networks can be locality sensitive hash functions

Deep Double Descent: Where bigger models and more data hurt

SGD Learns Functions of Increasing Complexity

Robust Influence Maximization for Hyperparametric Models

Robust Neural Networks are More Interpretable for Genomics