MV

Michal Valko

Bulding something new · Research at Inria, Lectures at ENS/MVA · Ex: Llama at GenAI, Meta; Gemini and BYOL at Google DeepMind

San Francisco, California

Overview 

Michal Valko is a Senior Staff Research Scientist at Google DeepMind, known for his work on projects like Llama at Meta and BYOL at Google DeepMind. With a background in reinforcement learning and machine learning, he has also held roles at Inria and École normale supérieure Paris-Saclay as a Junior Scientist and External Lecturer.

Work Experience 

  • Chief Models Officer, Member of the Founding Team, Member of Technical Staff

    2025 - Current

    Developing the next generation of language models

  • Principal Llama Engineer

    2024 - 2024

    new policy-gradient algorithms for Llama 3 and 4 post-training, improving math, reasoning, and code skills

Meta is a social technology company that enables people to connect, find communities, and grow businesses.

Raised $25,607,817,488.00 from ValueAct Capital.

  • Senior Staff Research Scientist

    2021 - 2024

    reinforcement learning with human feedback, online NashMD learning and offline IPO alignment for Gemini 2

  • Staff Research Scientist

    2019 - 2021

    BYOL for self-supervised learning, BYOL-Explore & BYOL-Hindsight for world models, BGRL for SSL graph embeddings

  • External Lecturer (CEV)

    2014 - 2023

    Teaching the graduate course "Graphs in Machine Learning" for the Master 2 program MVA - Mathématiques / Vision / Apprentissage.

  • Experienced Junior Scientist

    2013 - 2023

    new Monte-Carlo tree search techniques, solving open problem in online kernel and graph sparsification

  • Junior Scientist

    2012 - 2013

    Researcher in SequeL team at Inria Lille - Nord Europe, France, lead by Philippe Preux and Rémi Munos. Designing algorithms that would require as little human supervision as possible.

  • postdoctoral researcher

    2011 - 2012

    bandits, inverse reinforcement learning. Semi-supervised learning. composing learning for artificial cognitive systems

  • Research Assistant

    2006 - 2011

    Statistical anomaly detection methods for identification of unusual outcomes and patient management decisions. I combined max--margin learning with distance learned to create and anomaly detector, which outperforms the hospital rule for Heparin Induced Thrombocytopenia detection. I later scaled the system for 5K patients with 9K features and 743 clinical decisions per day. At the recent study, from 222 alerts 50\% were highly relevant. Mass-spec: I built a framework for the cancer prediction from high--throughput proteomic and genomic data sources. I found a way to merge heterogeneous data sources: My fusion model was able to predict pancreatic cancer from Luminex combined with SELDI with 91.2% accuracy.

  • Teaching Assistant

    2005 - 2005

    I was a TA for "Introduction of Programming" for Prof. Wizzard. Giving recitations, helping students and grading. http://www.cs.pitt.edu/~michal/hp/ta2061-0007

  • Graduate Research Intern

    2010 - 2010

    Large Scale Semi-Supervised Learning. Multi-manifold learning. I parallelized online harmonic solver to process 1 TB of video data in a day. I am working on the multi-manifold learning that can overcome changes in distribution. I am showing how the online learner adapts as to characters' aging over 10 years period in Married ... with Children sitcom. The research was part of Everyday Sensing and Perception (ESP) project.

An ecosystem of software & hardware vendors, integrators and solution providers focused on adoption of NFV and SDN-based solutions.

  • Intel Research Grad Intern

    2009 - 2009

    I extended graph-based semi-supervised learning to the structured case and demonstrated on handwriting recognition and object detection from video streams. Regularized harmonic function solution: The algorithm outputs a confidence of inference and uses it for learning. I came up with an online algorithm that on the real-world datasets recognizes faces at 80-90% precision with 90% recall.

An ecosystem of software & hardware vendors, integrators and solution providers focused on adoption of NFV and SDN-based solutions.

  • Research Assistant

    2003 - 2006

    Computer modeling of plausible neural network systems: I was modelling basic learning function at the level of synapses. I designed a model that is able to adapt to the regular frequencies with different a rate as the time flows. I used genetic programming to find biologically plausible networks that distinguish different gamma distribution and provided explanation of the strategies evolved.

Articles About Michal

Relevant Websites