Latest move
JetBrains → destination not stated
Language model pre-training researcher, now on Cohere's Foundations team in Berlin. Led JetBrains' Mellum code models for two and a half years, including Mellum2 pre-training; earlier did crowdsourcing, evaluation and RLHF research at Yandex and Toloka AI. BSc and MSc from MIPT.
Path
CohereMember of Technical Staff, Foundations2026- JBJetBrainsStaff Research Engineer / Team Lead (Mellum)2024–2026
- YTYandex / Toloka AIAI Research Scientist2020–2024
- MIMoscow Institute of Physics and TechnologyMSc, Computer Science2021–2023
- MIMoscow Institute of Physics and TechnologyBSc, Computer Science2017–2021
- MIMIPT & Inria NANO-DResearch Intern2020–2020
- YAYandexSoftware Engineering / ML Intern2018–2019
Papers
- Mellum2 Technical ReportarXiv 2026Describes Mellum 2, an open-weight 12B mixture-of-experts language model with 2.5B active parameters per token, built for software engineering tasks.
- Mellum: Production-Grade in-IDE Contextual Code Completion with Multi-File Project UnderstandingarXiv 2025Presents the Mellum family of 4B open-weight code completion models built for JetBrains IDEs, trained on about 4T tokens of permissively licensed code.
- Learning from Crowds with Crowd-KitJournal of Open Source Software 2021Presents Crowd-Kit, a Python toolkit of quality-control algorithms for crowdsourced labels, such as truth inference and learning from crowds.
SourcesLatest move: Xnikita-pavlichenko.comEducation and jobs are read from their own homepage or employer's page; papers checked on arXiv. From LinkedIn only the person's own public posts are read; GitHub is a link only.