Aleksandr Petiushko Александр Петюшко

Aleksandr Petiushko Александр Петюшко

Director, Head of ML Research / Adjunct Professor / PhD

Nuro

Sofia University

Biography

Principal R&D Researcher (15+ years of experience), R&D Technical Leader (10+ years of experience), and R&D Manager (8+ years of experience). Running and managing industrial research and academic collaboration (35+ publications, 30+ patents). Inspired by theoretical computer science and how it changes the world.

Dr. Aleksandr Petiushko is a Director, Head of Machine Learning Research at Autonomous Driving company Nuro (Mountain View, California), an Adjunct Professor at Sofia University (Palo Alto, California) teaching courses on ML and AI, and a lecturer at Lomonosov MSU and MIPT, giving lectures on the Theory of Deep Learning. Before Nuro, worked as a Team Lead / Scientific Expert, Chief Scientist at Huawei, as a Managing Director / Leading Scientific Researcher at Artificial Intelligence Research Institute. The Ph.D. dissertation is at the intersection of Discrete Mathematics and Computer Linguistics. Research interests lie in the applications of empirical and theoretical robustness (publications at ECCV, IJCAI, AAAI, CVPR, NeurIPS).

Interests
  • Autonomous Driving
  • Deep Learning
  • Robustness Theory
  • Discrete Mathematics
Education
  • PhD in Theoretical Computer Science, 2016

    Lomonosov Moscow State University

  • MSc in Mathematics, 2006

    Lomonosov Moscow State University

Skills

Research
Leadership
Lecturing

Industry Experience

 
 
 
 
 
Nuro
Director, Head of ML Research
February 2022 – Present California, USA

Responsibilities:

  • Managing and hiring the team of highly skilled diverse talents in ML (10+ FTEs, dotted line reports, consultants, interns)
  • Internal cross-collaboration and ideas brainstorming with other Behavior and Perception teams
  • Technical Roadmaps
  • State-of-the-Art frontier research
  • Academia collaboration

Technical Directions:

  • Prediction (including conditional and joint), Planning (including Diffusion-based), and Motion Selection (RL-based)
  • Closed-loop Reasoning
  • Synthetic Data Generation
  • Reward and Issue Predictor models
  • Agent-centric and scene-centric encoders
  • Scaling laws in Behavior
  • Robustness and Uncertainty of Autonomy Stack
  • LLM, VLM, Perception-Behavior interface

Other achievements:

  • Internal FTEs ML education through a series of Deep Learning Theory lectures
  • Main ideologist and program owner of Nuro ML University (100+ FTEs involved)
  • Created Nuro Tech Talks series (top robotics researchers share their ideas for Nuro)
  • Made SotA exploration a working internal pipeline
  • Constantly increasing Nuro’s visibility (participation at conferences, workshops, seminars, etc)
 
 
 
 
 
Artificial Intelligence Research Institute
Managing Director, Leading Scientific Researcher
September 2021 – February 2022 Moscow, Russia

Fusion Brain Research Director.

Responsibilities:

  • Research roadmap formulation
  • Research team hiring
  • Leading the research team (6 people)
  • Academia collaboration

Technical Directions:

  • Multi-modality
  • Multi-tasking
  • Retrieval-based systems
 
 
 
 
 
Huawei Research Center
Chief Scientist, Team Leader
September 2014 – September 2021 Moscow, Russia

Technical Lead and Research Manager.

Responsibilities:

  • Video Intelligence direct management (~15 people, plus interns)
  • Fundamental Research (~25 experts from different groups) team leadership
  • Academia collaboration projects on Fundamental Research in AI
  • Interviewing young talents
  • Mentoring interns
  • Remote work coordination

Technical Directions:

  • Video Recognition and Retrieval (MML, MDMMT, top place on smth2smth)
  • Trajectory Prediction (top places on Argoverse and Apolloscape benchmarks)
  • Face Recognition (Detection and Identification)
  • Domain Adaptation (computer vision)
  • Multi-task end2end (keypoints, detection, segmentation)
  • Storage (cold storage, erasure coding, object distribution, file type recognition with ML)
  • NeuralNets for Computer Vision (Super Resolution, DeBlur, Compression)
  • Robustness in DL
  • Geometrical DL

Other achievements:

  • Organized and maintained the Fundamental Research team (with publishing activity)
  • Organized and initially led the first Neural Net direction at Huawei MRC
  • Proposed and prepared the multi-year funding project Fundamental Research in AI at Huawei MRC
 
 
 
 
 
LSI Corporation (now Broadcom)
Senior Software Engineer
November 2008 – May 2014 California, USA (remote)

Engineering responsibilities:

  • Media gateway (DSP for audio)
  • WCDMA protocol (preamble detection)
  • LDPC (trapping set analysis)
  • Gesture recongintion (classical ML in CV)
  • Flash controller (wear leveler, garbage collector)

Other achievements:

  • 30+ patents on storage systems, ML and Computer Vision
 
 
 
 
 
RIRC of Briansk region
Software Engineer
February 2007 – November 2008 Briansk, Russia
Creating and supporting of a large DB (~1-10M records) on Oracle, implementing its functionality using PL/SQL; a number of client interfaces (Perl – CGI\Tk, AJAX (prototype, jQuery), Delphi) for interconnection with DB; Windows SVN server administration.
 
 
 
 
 
Computer Technologies
Software Engineer
September 2006 – April 2007 Briansk, Russia
Creating and supporting of bundled software to automate Internet provider tasks (accounting systems, billings, client and server-side for payment systems CyberPlat and OSMP using MySQL, PostgreSQL and Perl); implementing different web-interfaces and admin scripts (Perl, bash + sed, grep, awk) on demands for the administration department.
 
 
 
 
 
Auriga
Integration and testing department specialist
February 2006 – June 2006 Moscow, Russia
RTOS lynxOS testing in different environments, different scripts for testing (Perl, bash + sed, grep, awk); participating in porting utilities, OS code bug fixing (C); using CVS, RCS.
 
 
 
 
 
Neurocom
Software Engineer
December 2003 – June 2006 Moscow, Russia
Speech experiments, algorithms development (Matlab, C) and speech recognition in noisy conditions.
 
 
 
 
 
Optical and Electronic technologies
Software Engineer
Optical and Electronic technologies
July 2005 – April 2006 Moscow, Russia
Statistical experiments, algorithms development using Matlab for observation and tracing problem (video data from infrared imager).

Awards

Huawei Research Center
Best Mentor
Huawei Research Center
Fundamental Research Incentive
Huawei Research Center
Team Golden Medal
Huawei Research Center
Outstanding Individual
Huawei Research Center
Future Star
Huawei Research Center
Team Golden Medal
Huawei Research Center
Top Selling Point Delivery
Huawei Research Center
Technology Innovation and Breakthrough
Huawei Research Center
Quality Star
Huawei Research Center
Golden Luban
Huawei Research Center
Excellent Delivery and Customer Success
Huawei Research Center
Outstanding Contractor

Academia Experience

 
 
 
 
 
Sofia University
Adjunct Professor
July 2023 – Present California, USA
Machine Learning and Artificial Intelligence courses.
 
 
 
 
 
Moscow Institute of Physics and Technology (MIPT)
Visiting Professor
February 2023 – Present Moscow, Russia
Theory of Deep Learning courses.
 
 
 
 
 
Lomonosov Moscow State University (Lomonosov MSU)
Visiting Professor
February 2019 – Present Moscow, Russia
Python, Computer Vision, Machine Learning, Theory of Deep Learning courses.

Academic Service

 
 
 
 
 
Conference Reviews
April 2023 – Present
Serving as a reviewer for publications at conferences NeurIPS, ICLR, CVPR, ICML.
 
 
 
 
 
Journal Reviews
May 2021 – Present
Serving as a reviewer for publications in Neurocomputing.

Courses taught

Sofia University
Machine Learning
Sofia University
Machine Learning
Sofia University
Machine Learning
Sofia University
Machine Learning
Together with I.L.Mazurenko
Description of the course
Together with D.N.Babin and I.E.Ivanov
Together with D.N.Babin, I.E.Ivanov, A.S.Ivanyuta and S.P.Korviakov
Together with I.E.Ivanov
Together with D.N.Babin and I.E.Ivanov
Together with D.N.Babin, I.E.Ivanov, A.S.Ivanyuta and S.P.Korviakov
Together with D.N.Babin and I.E.Ivanov
Together with I.E.Ivanov
Together with D.N.Babin and I.E.Ivanov
Together with D.N.Babin and I.E.Ivanov

Education

 
 
 
 
 
Lomonosov Moscow State University (Lomonosov MSU)
Doctor of Philosophy (Ph.D.)
March 2016 – March 2016 Moscow, Russia

Russian name - “Candidate of Physical and Mathematical Sciences”, defended at Lomonosov MSU, Department of Mechanics and Mathematics.

Dissertation “Bigram Languages”, major 01.01.09 - Discrete Mathematics and Mathematical Cybernetics. Abstract. Youtube recording (all links are in Russian).

 
 
 
 
 
Lomonosov Moscow State University (Lomonosov MSU)
Postgraduate Student
September 2009 – November 2012 Moscow, Russia
Postgraduate study at Lomonosov MSU, Department of Mechanics and Mathematics. Major 01.01.09 - Discrete Mathematics and Mathematical Cybernetics.
 
 
 
 
 
Lomonosov Moscow State University (Lomonosov MSU)
Master of Science (M.Sc.)
September 2000 – July 2006 Moscow, Russia

Russian name - “specialist” (theoretical and applied mathematics), major in the discrete mathematics at Lomonosov MSU, Department of Mechanics and Mathematics.

Thesis “Dynamic adjustment of signals”. Incomplete (but the best that I could find) text (in Russian).

GPA 4.89/5, Diploma with honors.

 
 
 
 
 
Briansk Pushkin's Lycee
High School
January 1997 – July 2000 Briansk, Russia

High School, Briansk Pushkin’s Lycee, Physics and Mathematics major.

Gold medal.

Recent Publications

Multi-Constraint Safe RL with Objective Suppression for Safety-Critical Applications
Mutual modality learning for video action classification
Nonparametric Uncertainty Quantification for Single Deterministic Neural Network
Quadric Hypersurface Intersection for Manifold Learning in Feature Space

Projects

*
Nuro Tech Talks
Organizing a company-wide technical talk series.
School of Huawei Advanced Research Education
Main technical coordinator and the owner of Machine Learning and Computer Vision specialization.
Hackathon Metric Learning for facial descriptors
Hackathon with the main goal to improve the FaceID by utilizing the metric learning methods.
Python imresize
Bit-exact copy of a MatLab function called imresize().
Regex Calculator
A simple calculator built using only regular expressions.

Personal Blog

Contact

I’m open to any public and private technical talks and collaboration regarding Autonomy, Self-Driving, Certified and Adversarial Robustness, and Discrete Mathematics; interested to share the methodological principles of teaching Machine Learning.