# Timur Kharisov

> Timur Kharisov, PhD researcher at the University of Basel. Optimization for deep learning, training dynamics, stochastic and distributed optimization.

This is Timur Kharisov's personal academic website. Navigation uses game-menu labels; the linked pages below contain ordinary server-rendered text and work without JavaScript.

## Profile

I am a PhD student at the [University of Basel](https://www.unibas.ch/en.html), specializing in Optimization for Deep Learning within the [OMLS team](https://sites.google.com/view/omls-group) under the supervision of [Prof. Aurelien Lucchi](https://scholar.google.com/citations?user=V1ONSgIAAAAJ&hl=en). My research interests encompass the training dynamics of deep learning models, stochastic optimization, and distributed optimization. Previously, I graduated from the [M2 Optimization](https://www.master-in-optimization.fr/) program at Université Paris-Saclay as a [Sophie Germain](https://fondation-hadamard.fr/en/our-programs/transversal-programs/graduate-program/apply-for-a-sophie-germain-scholarship/) FMJH scholar. I hold a BSc in Applied Mathematics and Computer Science from [MIPT](https://en.wikipedia.org/wiki/Moscow_Institute_of_Physics_and_Technology) (Moscow) and have completed the [YSDA](https://dataschool.yandex.com/) Data Science track. I have gained research experience through an internship in the [INRIA Fairplay](https://team.inria.fr/fairplay/) group, an internship in [Prof. Peter Richtarik’s](https://richtarik.org/) group at KAUST, a quantitative research internship at [WorldQuant](https://www.worldquant.com/), and a machine learning internship at [Yandex Alice](https://en.wikipedia.org/wiki/Alice_(virtual_assistant)).

## Pages

- [Character profile](https://kharisov.me/): The person behind the gradients
- [Campaign](https://kharisov.me/projects/): Projects & experiments
- [Intel](https://kharisov.me/research/): Research, papers & talks
- [Patch notes](https://kharisov.me/news/): News & recent updates
- [World map](https://kharisov.me/location/): A somewhat nomadic backstory
- [Comms](https://kharisov.me/contact/): Let's start a conversation

- [Swiss Optimization Symposium talk slides](https://kharisov.me/sos-26/): What has Edge of Stability to do with the Convergence Theory? August 2026. Timur Kharisov, Foivos Alimisis, and Aurelien Lucchi.
- [Sitemap](https://kharisov.me/sitemap.xml)

## Research context

The Intel page pairs a schematic river valley with sharpness curves from the talk. The valley is not a measured parameter trajectory. Update-averaged sharpness means curvature averaged along a finite optimization update, not a moving average over training time.

## Public profiles

- [Google Scholar](https://scholar.google.com/citations?user=qWasd7MAAAAJ&hl=en)
- [GitHub](https://github.com/tkharisov7)
- [LinkedIn](https://www.linkedin.com/in/tkharisov7/)