I am a senior research scientist at NVIDIA's Applied Deep Learning Research team. Before I was at Isomorphic Labs working on AI-first drug discovery. Prior to that, I was a PhD student at the University of Oxford working on multitask reinforcement learning with graph-based state-action representation. My scientific advisor is Shimon Whiteson. I'm also known as yobibyte. Ping me if you are interested in my work. I also have a telegram channel where I share links on machine learning, software engineering, music and other random stuff.

industry

• NVIDIA, Senior Research Scientist, June 2024/Now

• Isomorphic Labs, London, UK, Senior Research Scientist, November 2022/June 2024

  Leading a molecular property prediction project focusing on small molecules.

• DeepMind, London, UK, Research Intern, May 2022/September 2022

• Microsoft Research, Cambridge, UK, Research Intern, January 2021/March 2021

  Working with Ryota Tomioka and Andrew Fitzgibbon.

• Facebook AI Research, London, UK, Research Intern, September 2020/December 2020.

  Working with Tim Rocktaeschel on NetHack learning environment.

• NVIDIA, Santa Clara, US, Research Intern, June 2019/September 2019
  • Part of the Applied Deep Learning Research team, working on Reinforcement Learning applications for SAT solvers. Paper got accepted to NeurIPS 2020.

• Latent Logic, Oxford, UK (acquired by Waymo), Junior Researcher, October 2017/September 2018
  • Working on Human Behaviour Modelling using Reinforcement Learning and Learning from Demonstration techniques.
  • Behaviour Models for Autonomous Vehicle Simulators patent.
  • "Learning from Demonstrations in the Wild", ICRA 2019

• Microsoft Research Cambridge, UK, visiting student, March 2017/May 2017

  Working with Sebastian Nowozin and Katja Hofmann on my master thesis.

• RWTH Computer Vision Group, student assistant, October 2015/December 2016
  • Implementing the ROS node for head orientation prediction.
  • Working on human re-identification.

• Terminal, Java EE Developer/Project Manager, November 2013/August 2013

  Leading a team of four people developing a warehouse management system from scratch. Currently the system is implemented (~20 users).

• CDC.ru, Android developer, May 2013/November 2013

  Member of a development team of a mobile app for an ERP system. Completely refactored reports module and made a large amount of bug fixes.

• Imagespark, Java/Android developer, January 2013/May 2013

academia

Publications&Preprints

• In Defense of the Unitary Scalarization for Deep Multi-Task Learning, Vitaly Kurin (equal contribution, Alessandro De Palma (equal contribution), Ilya Kostrikov, Shimon Whiteson, M. Pawan Kumar, NeurIPS 2022

• MiniHack the Planet: A Sandbox for Open-Ended Reinforcement Learning Research, Mikayel Samvelyan, Robert Kirk, Vitaly Kurin, Jack Parker-Holder, Minqi Jiang, Eric Hambro, Fabio Petroni, Heinrich Kuttler, Edward Grefenstette, Tim Rocktaeschel, Datasets and Benchmarks track, NeurIPS 2021

• Snowflake: Scaling GNNs to High-Dimensional Continuous Control via Parameter Freezing, Charlie Blake, Vitaly Kurin, Maximilian Igl, Shimon Whiteson, NeurIPS 2021

• My Body is a Cage: the Role of Morphology in Graph-Based Incompatible Control, Vitaly Kurin, Maximilian Igl, Tim Rocktaeschel, Wendelin Boehmer, Shimon Whiteson, ICLR 2021.

• Can Q-Learning with Graph Networks Learn a Generalizable Branching Heuristic for a SAT Solver?, Vitaly Kurin, Saad Godil, Shimon Whiteson, Bryan Catanzaro, NeurIPS 2020.

• Deep Coordination Graphs, Wendelin Boehmer, Vitaly Kurin, Shimon Whiteson, ICML 2020.

• Fast Efficient Hyperparameter Tuning for Policy Gradients, Supratik Paul, Vitaly Kurin, Shimon Whiteson, NeurIPS 2019.

• Fast Context Adaptation via Meta-Learning, Zintgraf, Luisa M., Kyriacos Shiarlis, Vitaly Kurin, Katja Hofmann, and Shimon Whiteson, International Conference on Machine Learning (ICML) 2019.

• Learning from Demonstration in the Wild, Feryal Behbahani, Kyriacos Shiarlis, Xi Chen, Vitaly Kurin, Sudhanshu Kasewa, Ciprian Stirbu, João Gomes, Supratik Paul, Frans A. Oliehoek, Joao Messias, Shimon Whiteson, ICRA 2019.

• The Atari Grand Challenge Dataset, Vitaly Kurin, Sebastian Nowozin, Katja Hofmann, Lucas Beyer, and Bastian Leibe, arXiv preprint arXiv:1705.10998, 2017.

• Towards a Principled Integration of Multi-Camera Re-Identification and Tracking through Optimal Bayes Filters, Lucas Beyer, Stefan Breuers, Vitaly Kurin, Bastian Leibe, CVPR 2017 ReID and MTMCT workshop.

places

• University of Oxford, DPhil (PhD) in Computer Science, 2018-2023
  • Supervisor: Shimon Whiteson
  • I am working on applications of Graph Neural Networks in Reinforcement Learning.

• RWTH Aachen University, M.Sc. 2015-2017 (Media Informatics program)
  • It was a joint program between University of Bonn and RWTH Aachen, I spent a year in each of them. I did a thesis on Imitation Learning.
  • Courses
    • Game AI
    • High-Performance Matrix computations
    • Pattern Recognition
    • Deep Learning for Multimedia Pattern Recognition
    • Advanced Machine Learning
    • Distributed Systems
    • Computer Graphics
  • Seminars&Labs
    • The report on Neural Turing Machines for "Selected Topics in Human Language Technology and Pattern Recognition" seminar.
    • The SS16 GameAI lab report with prof. Bauckhage where we played PlanetWars with Q-learning.

• Moscow State University, B.Sc. 2012-2015 (Faculty of Computational Mathematics and Cybernetics)
  • Real analysis, Complex analysis
  • Algorithms Algorithmic languages
  • Algebra and Analytical Geometry
  • Functional Analysis
  • Discrete Mathematics
  • Computer Architecture and Assembly language, Operating Systems
  • Differential Equations
  • Equations of Mathematical Physics
  • Probability Theory and Mathematical statistics
  • Programming languages
  • Game Theory and Operational Research
  • Optimal Control
  • Databases
  • Computer Graphics
  • Computer Networks
  • Numerical Methods

• Moscow State Institute of International Relations, B.A .2008-2012 (Faculty of Economics, Informational Systems in International Business program)
  • Mathematical Analysis
  • Linear Algebra
  • Probability theory and Mathematical statistics
  • Econometrics

• Misc
  • Prof. Andrew Ng ML course on coursera
  • Prof. Hinton course on coursera

Teaching

• Reinforcement Learning, University of Oxford, Michaelmas 2021, Teaching Assistant

• "Leading Data Science course", Center for Digital Technology and Management, invited lecturer, Munich, Summer 2021:
  • Graph Neural Networks
  • Reinforcement Learning

• Machine Learning, University of Oxford, Michaelmas 2018, Marker.

• Artificial Intelligence, University of Oxford, Hilary 2019, Marker.

Supervision

• Jeffrey Mak, master thesis on multiagent RL, 2019, co-supervision with Wendelin Boehmer

• Charles Blake, master thesis on scaling GNNs in RL, 2020, co-supervision with Max Igl, Submitted to NeurIPS 2021

🏆Prizes and Scholarships

• Outstanding Reviewer at ICLR 2024.
• United Kingdom Global Talent Visa
• NeurIPS 2022 Scholarship Award
• Outstanding Reviewer at ICLR 2021.
• Samsung R&D Institute UK studentship, 2018-2023
• NeurIPS 2019 Travel Award

Service

• NetHack NeurIPS 2021 Challenge, organizer.
• GNNSys'21 — Workshop on Graph Neural Networks and Systems (in conjunction with MLSys 2021), program committee.
• ICLR 2021, 2022, 2023 reviewer.
• ICML 2021, 2022, reviewer.
• ICML 2019 Imitation, Intent, and Interaction (I3) workshop, reviewer.
• Robotics and Autonomous Systems CDT Conference 2019, organiser, reviewer.
• NeurIPS 2021, 2022, reviewer
• COLLAs 2022, program committe
• GroundedML: Workshop on Anchoring Machine Learning in Classical Algorithmic Theory, ICLR 2022, reviewer
• Competition Track NeurIPS 2022, reviewing proposals
• Leading Deep Learning Reading Group at the University of Oxford, 2018-2022

Talks

• Tutorial on Graph Neural Networks, ICAIF'21 Workshop on NLP and Network Analysis in Financial Applications

“My Body is a Cage: the Role of Morphology in Graph-Based Incompatible Control”, CDT 2021 Student Conference, August 2021

"Incompatible Multitask Reinforcement Learning", ELLIS Interactive Learning and Interventional Representations (ILIR) Workshop, June 21th, 2021

"My Body is a Cage: the Role of Morphology in Graph-Based Incompatible Control", Valence Discovery Graph Journal Club, June 3rd 2021

"Graph Neural Networks for Reinforcement Learning, Machine Learning Forum at CERN, 27 May, 2021.

• "My Body is a Cage: the Role of Morphology in Graph-Based Incompatible Control", Berkeley MARL Seminar, 19 March 2021

• My Body is a Cage: the Role of Morphology in Graph-Based Incompatible Control", RL Seminar, University of Edinburgh, 11 February 2021

• ”Graph Networks” tutorial, at the Robotics and Autonomous Systems CDT Conference 2019, Edinburgh, 3-4 June 2019

• ”Graph Networks”, at the Centre for Visual Information Technology (CVIT) at International Institute of Information Technology, Hyderabad, India, March 19th 2019.

• ”Learning from Demonstration” at Databeers Branded Edition with BBVA Data & Analytics, September 3rd 2018.

Summer Schools

• London Geometry and Machine Learning Summer School, 2021

• Oxford Summer School in Economic Networks, 2021

• Machine Learning Summer School, Madrid 2018

tools

Tools

• Linux
  • I've been using linux for more than 15 years already.
  • I'm using Arch Linux + i3 window manager. I like how fast and responsive i3 is and I enjoy not using my mouse. I've been using a lot of other distros throughout my life: Ubuntu/Kubuntu, Suse, Debian, Mint, Fedora. I tried to install Gentoo back in 2006, but I could not make the pesky video driver work on my Sony Vaio laptop.

• Most useful tools in my toolbox:
  • terminal (I use kitty)
  • ssh
  • ag-silversearcher
  • neovim
  • jupyter notebooks for visualisation mostly
  • sacred/wandb for the experiment management
  • git for version control

• 🧪Deep Learning frameworks
  • I mostly use Jax nowadays.
  • I've been working with Pytorch, Theano, Tensorflow and Chainer throughout different stages of my career.

• Remote servises
  • I've been using AWS and Azure as well as private companies clusters for running my experiments.
  • I love slurm (when it's set up)

• Programming Languages
  • I mostly use python nowadays, but I've used a bunch of other languages throughout my career (see below).
  • I've been a Java software engineer (SE&EE) at the beginning of my career, when Java 8 appeared and people got excited about lambda expressions there. I developed for Android a bit.
  • I used mySQL when I was working as a Java developer. I used noSQL (graph) databases (neo4j) for my undergrad thesis.
  • I've been playing with other languages for fun and university courses:
    • C>
    • C++
    • Assembly>
    • Lisp
    • I started learning Rust recently.

• Human Languages
  • Russian (native)
  • English (full professional profficiency)
  • Chinese (beginner, HSK level 4)
  • French (beginner)
  • German (beginner)

misc

• I love mathematical puzzles, this is my project euler profile.

  

• I am learning chess. My current rapid rating is about 1800. This is my chess.com profile.

• I love music. I can play the piano and an electric guitar. I started learning drums a year ago.

• I am a proud parent of a 9yo son. I love soldering different gadgets and playing football and chess with him.