Igor Adamski

Igor Adamski

About

Detail

Co-Founder
United Kingdom

Timeline

work
Job
school
Education
folder
Project
auto_stories
Publication

Résumé

Jobs verified_user 0% verified
  • Alaro
    Co-Founder
    Alaro
    Apr 2025 - Current (1 year 3 months)
  • Monzo Bank
    Machine Learning Engineer
    Monzo Bank
    Jul 2021 - Aug 2021 (2 months)
  • AGITProp
    AI Researcher
    AGITProp
    Jun 2024 - May 2025 (1 year)
  • Monzo Bank
    Business Intelligence Developer
    Monzo Bank
    Feb 2021 - Jul 2021 (6 months)
  • Algopolis
    Lead Quantitative Researcher
    Algopolis
    Nov 2022 - Apr 2024 (1 year 6 months)
  • Algopolis
    Quantitative Researcher & Developer
    Algopolis
    Aug 2021 - Jan 2023 (1 year 6 months)
  • Five AI
    Machine Learning Engineer
    Five AI
    Sep 2020 - Dec 2020 (4 months)
  • Gardenia Technologies
    Data Scientist
    Gardenia Technologies
    Sep 2018 - Sep 2019 (1 year 1 month)
  • deepsenseai
    Deep Learning Research Intern
    deepsenseai
    Jul 2017 - Oct 2017 (4 months)
Education verified_user 0% verified
  • University of Cambridge
    University of Cambridge
    University of Cambridge
    Jan 2019 - Dec 2020 (2 years)
  • Imperial College London
    Imperial College London
    Imperial College London
    Jan 2015 - Dec 2018 (4 years)
  • International baccalaureate
    International Baccalaureate
    International baccalaureate
    Jan 2012 - Dec 2015 (4 years)
    Received 44/45 points. Subjects: Higher level: Mathematics (7), Chemistry (7), English B (7) Standard level: Economics (7), Geography (7), Polish A (6)
Projects (professional or personal) verified_user 0% verified
    Publications verified_user 0% verified
    • I
      Distributed Deep Reinforcement Learning: Learn how to play Atari games in 20 minutes
      ISC High Performance Frankfurt Conference Proceedings Springer
      Jun 2018
      We present a study in Distributed Deep Reinforcement Learning (DDRL) focused on scalability of a state-of-the-art Deep Reinforcement Learning algorithm known as Batch Asynchronous Advantage Actor Critic (BA3C). We show that using the Adam optimization algorithm with a batch size of up to 2048 is a viable choice for carrying out large scale machine learning computations. This, combined with careful reex- amination of the optimizer’s hyperparameters, using synchronous train- ing on the node level (while keeping the local, single node part of the algorithm asynchronous) and minimizing the model’s memory footprint, allowed us to achieve linear scaling for up to 64 CPU nodes. This corre- sponds to a training time of 21 minutes on 768 CPU cores,
    This is a community-created genome.