Ekaterina Lobacheva
Postdoctoral Research Fellow at Mila - Quebec Artificial Intelligence Institute
Tbilisi, Canada
Summary
👤
Senior🎓
Top SchoolEkaterina Lobacheva is a deep learning researcher and postdoctoral fellow at Mila and Université de Montréal with nine years of experience studying how optimization and training dynamics shape learned representations and generalization. Her work—published at NeurIPS, AAAI, EMNLP and ICCV—spans loss landscapes, mode connectivity, ensembling in transfer learning, and effects of large initial learning rates on training. She collaborates closely with Bayesgroup researchers and advisors Nicolas Le Roux and Irina Rish, investigating gradient opposition and its impact on training large foundational models. Previously she led research and taught machine learning and Bayesian methods at the Higher School of Economics and contributed applied work at Kaspersky and Western University. Beyond publications, she combines theoretical analysis with empirical probing of training trajectories, often connecting function-space perspectives to practical ensembling and transfer strategies. Based in Tbilisi/Canada, she maintains an active research homepage and CV showcasing reproducible results and cross-group collaborations.
9 years of coding experience
PhD in Computer Science, Computer Science, PhD in Computer Science, Computer Science at Higher School of EconomicsComputer Science, Computer Science at Yandex School of Data AnalysisLomonosov Moscow State University
Github Skills (4)
python6
deep-learning5
machine-learning5
tensorflow4
Programming languages (1)
Github contributions (5)
bayesgroup/HSE_ML_research_seminar
Sep 2019 - May 2022
Contributions:126 commits, 116 pushes, 1 branch in 2 years 8 months
bayesgroup/bayesgroup.github.io
Sep 2016 - Jun 2019
Contributions:246 commits, 1 PR, 233 pushes in 2 years 9 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.