Edward Grefenstette
Director Of Research, Foundational Research Board Member, And Program Area Lead
Stony Stratford, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolEdward Grefenstette is a research leader and practitioner in machine learning and computational linguistics with 12 years of experience bridging foundational theory and applied systems. Currently a Director of Research and Program Area Lead at Google DeepMind and an Honorary Professor at UCL, he has steered RL and foundational ML efforts at top labs including DeepMind, Facebook AI, and Cohere. His work spans mathematical logic, information theory and language, and he contributes to practical ML tooling—evidenced by notable contributions to the PyTorch higher library for higher-order gradients and differentiable optimizers. Trained with a DPhil in Computer Science from Oxford and an MLitt in Philosophy, he uniquely combines formal rigor with engineering impact, moving ideas from proofs to production-scale research. Colleagues know him for making theoretically subtle concepts operational and for leadership that connects deep research to real-world model optimization.
12 years of coding experience
14 years of employment as a software developer
DPhil Computer Science, DPhil Computer Science at University of OxfordMLitt Philosophy, MLitt Philosophy at University of St AndrewsBSc Physics and Philosophy, BSc Physics and Philosophy at The University of Sheffield
English, French, Japanese, Italian
Stackoverflow
603reputation
28kreached
4answers
12questions
Github Skills (19)
algorithm10
pytorch10
python10
optimizers10
machine-learning10
deep-learning10
optimisation10
optimization10
unit-testing9
n9
rnn-model9
html6
macos6
database6
clang6
Programming languages (9)
Github contributions (5)
facebookresearch/higher
Oct 2019 - Oct 2020
higher is a pytorch library allowing users to obtain higher order gradients over losses spanning training loops rather than individual training steps.
Role in this project:
ML Engineer
ML EngineerContributions:3 reviews, 22 commits, 34 PRs in 1 year 1 month
Contributions summary:Edward primarily contributed to the `higher` library, focusing on improving and expanding its functionality for higher-order gradients in PyTorch. They addressed issues related to differentiable optimizers, particularly with frozen parameters, and introduced the capability to transform and manipulate gradients within the optimizers. Furthermore, they added support for patched RNN forward operations, with associated unit tests, which enhanced the scope and capabilities of the library in the context of deep learning model optimization. They also worked on refactoring and enhancing existing unit tests to ensure the library's robustness and feature set.
pytorchindividualloopsgradientstraining
egrefen/egrefen.github.io
Feb 2014 - Sep 2022
Contributions:79 commits, 30 pushes in 8 years 8 months
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.