Alex Ayoub

I am a Research Scientist at Google DeepMind in London, where I work on the science of reinforcement learning and post-training team. My research focuses on reinforcement learning theory, efficient reasoning in language models, and decision-focused machine learning systems.

Before joining Google DeepMind full-time, I was a PhD candidate in Computing Science at the University of Alberta, advised by the great Csaba Szepesvari and the great Dale Schuurmans.

Email / CV / Scholar / LinkedIn / GitHub

Research

I work on reinforcement learning and post-training, with a focus on efficient reasoning in language models and theory-driven algorithm design. My broader work includes sample-efficient reinforcement learning with function approximation and objective design for recommender systems.

ICLR 2026	Learning to Reason Efficiently with Discounted Reinforcement Learning A. Ayoub, K. Asadi, D. Schuurmans, C. Szepesvari, K. Bouyarmane ICLR, 2026
NeurIPS 2025	Eluder Dimension: Localise It! A. Bakhtiari, A. Ayoub, S. Robertson, D. Janz, C. Szepesvari NeurIPS, 2025 (Spotlight, top 3%)
RLC 2025	Rectifying Regression in Reinforcement Learning A. Ayoub, D. Szepesvari, A. Bakhtiari, C. Szepesvari, D. Schuurmans Reinforcement Learning Conference, 2025
WWW 2025	Does Weighting Improve Matrix Factorization for Recommender Systems? A. Ayoub, S. Robertson, D. Liang, H. Steck, N. Kallus The Web Conference, 2025
NeurIPS 2024	Almost Free: Self-Concordance in Natural Exponential Families and an Application to Bandits S. Liu, A. Ayoub, F. Sentenac, X. Tan, C. Szepesvari NeurIPS, 2024
RLC 2024	Mitigating the Curse of Horizon in Monte-Carlo Returns A. Ayoub, D. Szepesvari, F. Zanini, B. Chan, D. Gupta, B. Castro da Silva, D. Schuurmans Reinforcement Learning Conference, 2024
ICML 2024	Switching the Loss Reduces the Cost in Batch Reinforcement Learning A. Ayoub, K. Wang, V. Liu, S. Robertson, J. McInerney, D. Liang, N. Kallus, C. Szepesvari ICML, 2024
AISTATS 2024	Exploration via Linearly Perturbed Loss Minimisation D. Janz, S. Liu, A. Ayoub, C. Szepesvari AISTATS*, 2024 (Oral, top 1%)
NeurIPS 2023	Managing Temporal Resolution in Continuous Value Estimation: A Fundamental Trade-off Z. Zhang, J. Kirschner, J. Zhang, F. Zanini, A. Ayoub, D. Schuurmans NeurIPS, 2023
TMLR 2023	Resmax: An Alternative Soft-Greedy Operator for Reinforcement Learning E. Miahi, R. MacQueen, A. Ayoub, A. Masoumzadeh, M. White Transactions on Machine Learning Research, 2023
ICML 2021	Randomized Exploration for Reinforcement Learning with General Value Function Approximation H. Ishfaq, Q. Cui, V. Nguyen, A. Ayoub, Z. Yang, Z. Wang, D. Precup, L. F. Yang ICML, 2021
ICML 2020	Model-Based Reinforcement Learning with Value-Targeted Regression A. Ayoub, Z. Jia, C. Szepesvari, M. Wang, L. Yang ICML, 2020

Professional Experience

Google DeepMind, Research Scientist - Science of Reinforcement Learning and Post-Training (London, UK), 2026-Present

Google DeepMind, Student Researcher - Frontier AI Unit (London, UK), Aug 2025-2026

Education

University of Alberta, PhD in Computing Science, Sep 2021-May 2027 (Expected)

Supervisors: Csaba Szepesvari and Dale Schuurmans

University of Alberta, MSc in Computing Science, Sep 2019-Sep 2021

Thesis: Towards Sample Efficient Reinforcement Learning with Function Approximation (nominated for Outstanding Thesis Award)

Florida State University, BSc in Computational Science and Applied Mathematics, Jun 2015-May 2019

Service, Awards, and Skills

Awards and Service: Inaugural Netflix Graduate Research Fellowship; Co-organizer, RL Theory Workshop (2023, 2024); Workflow Chair, ICML 2022; University of Alberta Graduate Research Assistant Fellowship; Florida State University Research Assistant Grant; FSU President's List (2018).

Technical Skills: Python, C/C++, MATLAB, PyTorch, JAX, CuPy, Mathematica.

Teaching: Teaching Assistant, University of Alberta (Sep 2019-Present).

Website format adapted from Jon Barron's template.