Avatar

Egor Shulgin

PhD candidate

KAUST

Bio

I am a final-year PhD student at King Abdullah University of Science and Technology (KAUST), advised by Peter Richtárik. My research focuses on optimization for distributed machine learning and private federated learning.

During my PhD, I spent time at Apple and Samsung AI Center (Cambridge, UK), where I worked on heterogeneous, efficient, and personalized federated learning.

I am currently on the job market.

Apart from research, I am also passionate about hiking/backpacking, alpine skiing, and rationality.

Interests

  • Optimization
  • Deep Learning
  • Distributed Learning
  • Privacy and Security in ML

Education

  • BSc in Applied Mathematics, Computer Science and Physics, 2019

    Moscow Institute of Physics and Technology

Recent Posts

Workshop on Distributed Training in the Era of Large Models

Co-organized the KAUST Workshop on Distributed Training in the Era of Large Models.

Invited Talk at MBZUAI AI Speaker Series

Delivered an invited talk (recording) at the MBZUAI AI Speaker Series (Statistics and Data Science Seminar) in Abu Dhabi.

CEMSE Dean's List Award

Received the CEMSE Dean’s List Award for exceptional academic achievements at KAUST.

Research Visit at MBZUAI

Started a research visit at the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), hosted by Dr. Eduard Gorbunov.

Recognized as Expert Reviewer for TMLR

Recognized as an Expert Reviewer for Transactions on Machine Learning Research (TMLR).

Recent Publications

Quickly discover relevant content by filtering publications.

Beyond the Ideal: Analyzing the Inexact Muon Update

We provide analysis of the inexact orthogonalized update at Muon’s core, revealing a fundamental coupling between LMO inexactness …

Gluon: Making Muon & Scion Great Again! (Bridging Theory and Practice of LMO-based Optimizers for LLMs)

We propose Gluon, a new LMO-based optimizer with a refined generalized smoothness model that captures layer-wise geometry and closes …

Smoothed Normalization for Efficient Distributed Private Optimization

We design α-NormEC, the first differentially private distributed optimization algorithm with provable convergence guarantees for …

MAST: Model-Agnostic Sparsified Training

We introduce a novel optimization formulation incorporating pre-trained models and random sketch operators, enabling …

On the Convergence of DP-SGD with Adaptive Clipping

We provide the first comprehensive convergence analysis of SGD with quantile clipping, establishing theoretical guarantees for …