About Me
I’m mathematician that works on industry problems related to A/B testing, RecSys and AI/ML in general.
List of my contrtibutions:
– 6 conference publications (ICML, ICLR)
– 2 journal publications
– 4 patents
– 91 citations with h-index 5
– Contributor to open-source library CatBoost
Topics of research that I do:
– Learning-to-Rank (author of two state-of-the-art algorithms StochasticRank, YetiLoss)
– Diffusion processes (including simulating quantum mechanics, general theory of convergence of markov chains to diffusion processes, non-convex optimization using Langevin diffusion)
– Gradient Boosting (convergence and generalization in RKHS, non-convex optimization)
– Uncertainty estimation (bayesian, gaussian processes using gradient boosting)
– Non-convex optimization (for recsys)
Publications
Andrey Malinin, Liudmila Prokhorenkiva, Aleksei Ustimenko; ICLR 2021
Experience
ShareChat
Lead Machine Learning Scientist
July 2022 - Present
Leading Indian short video app with hundreds millions MAU.
Leading feed ranking modeling efforts and continuing researching science about theoretical foundations of gradient boosting, diffusion processes, learning-to-rank. Managing team of Machine Learning Enginiers to bring state-of-the-art approaches to our recsys. Leading efforts on bringing modern A/B testing techniques.
Yandex.Research
Research Scientist
September 2019 - May 2022
Yandex -- Leading Russian Internal Company (Search Engine, Taxi, Marketplace, etc). Yandex.Research -- lab focused on research.
Researching science about theoretical foundations of gradient boosting, diffusion processes, learning-to-rank. Consulting enginiering teams about how to apply my algorithms. Published 3 papers during my work there and 2 wrote 2 more papers that I’ve published later.
Yandex.Market
Data Scientist
April 2017 - August 2019
Largest Russian Marketplace (part of Yandex)
Researching science about theoretical foundations of gradient boosting, diffusion processes, learning-to-rank. Consulting enginiering teams about how to apply my algorithms. Invented algorithm StochasticRank that I’ve published later during work at Research.
Yandex.Research
Intern Research Scientist
May 2016 - February 2017
Leading Russian Internal Company (Search Engine, Taxi, Marketplace, etc).
Researching science about A/B testing and learning-to-select problem for Yandex.Market. As result of my work I’ve published my first ICML paper.
Education
Lomonosov Moscow State University
BA of Applied Mathematics and Computer Science.
2013-2017
Departamant of numerical methods of linear algebra.
Specialized on tensor decompositions, numerical linear algebra and algebraic geometry. Wrote my first 2 journal publications on linear algebra during my second year.
A Little More About Me
I’m mathematician with passion about Machine Learning and AI. Even though by background is theoretical and my papers are mostly theoretical I work closely with industry to close the gap between theory and practice by identifying problems of practitioners and implementing theoretically sound solutions to them.
I published 6 conference papers on tier-1 conferences (ICML, ICLR), 2 journal publications on linear algebra and hold 4 patents on my inventions.
I’m passionate about pushing forward theoretical understanding of AI using the most advanced methods of Stochastic Analysis.
All my theoretical breakthroughs have direct applications to industry, e.g. my state-of-the-art learning-to-rank algorithms, uncertainty estimation algorithms, gradient boosting.
I’m contributor to open source library CatBoost.
I’m proficient with C++, Python (including PyTorch and Tensorflow).
I help companies to find flaws in use of ML&Statistics, correctly formalize problems that they try to solve and propose best possible solutions, this includes A/B testing, RecSys, Ranking, Reinforcement Learning.
I have experience of technical leadership and management of teams of Machine Learning Enginiers of size up to 10.