Hayder Elesedy

I work as a Senior Machine Learning Researcher at Samsung researching LLM safety and differentially private, federated estimation.

My interests include reasoning, theoretical aspects of statistical machine learning, LLMs, deep learning and applications of machine learning to science.

I was a DPhil student in machine learning at Oxford, supervised by Varun Kanade and Yee Whye Teh.

I spent August-December 2022 at DeepMind working on U-Clip, supervised by Marcus Hutter.

I spent the second half of 2021 year doing AI research at X (formerly Google [x]) supervised by David Duvenaud.

In 2020 I was part of the Royal Society’s DELVE initiative. This is a multidisciplinary group formed in response to the coronavirus pandemic.

Before my PhD I was a quant at Tudor, and before that I studied maths and theoretical physics at Cambridge where I was fortunate to be taught by Stephen Siklos.

I was not a normal Cambridge applicant. Over the years I’ve helped a few others bridge the gap. If you want help, send me an email.

My work in the news

Note: I prefer my Arabic name Hayder, but prior to 2024 I used my Welsh name, Bryn.

Publications and Preprints

Projects