I am a research scientist and engineer with 4+ years of experience in recommender systems, with accepted papers at KDD, SIGIR, and RecSys. I am currently a Senior Research Scientist at AI VK, working in the research team led by Prof. Dr. Alexander D’yakonov, formerly the world’s #1 ranked competitor on Kaggle. There I work on long-term RL optimization for generative retrieval — I am the first author of Long-Term Optimization for Large-Scale Generative Retrieval with Off-Policy REINFORCE (KDD 2026) — and I research PLUM-style adaptation of LLMs to generative recommendation via Semantic IDs. I have also contributed to online A/B tests in VK Video that improved Total Viewing Time (TVT).

I hold a B.Sc. in Applied Mathematics and Computer Science from Lomonosov Moscow State University (MSU) and I am an M.Sc. candidate in Modern Computer Science at the Higher School of Economics (HSE), as well as a student of the selective Yandex School of Data Analysis (YSDA) program.

Previously, at Yandex, I led a three-person R&D team responsible for pretraining large-scale generative recommendation models, internal training-framework development, and inference optimization. I contributed to Sona, a single generative recommender that replaced the entire production cascade at Yandex Music (+4.53% Active Users, +6.30% Total Listening Time, +11.42% Likes in an online A/B test). I co-developed Argus, a billion-parameter generative recommender transformer that delivered strong online A/B gains across Yandex products (e.g. +2.26% Total Listening Time at Yandex Music, +1.7% GMV in e-commerce, +2% revenue in Ads ranking), and I introduced FSDP training to scale recommender transformers up to 7 billion parameters.

Research interests

Information retrieval, conversational recommendation, large language models, generative recommendation, in-text reasoning for recommendation, reinforcement learning, and the mathematical foundations of deep learning.

Find me

You can reach me by email or find me on Google Scholar, ORCID, GitHub, LinkedIn, Telegram, X / Twitter, Instagram, Kaggle, and Codeforces. My CV is available as a PDF.