Skander Moalla

I am considering full-time positions in Zurich, Paris, or London for Fall 2026. I bring a strong background in both research and engineering.

Hi! I’m a final-year PhD candidate in reinforcement learning (RL) and LLM post-training at EPFL, advised by Prof. Caglar Gulcehre (CLAIRE Lab).

I work on scaling reinforcement learning for LLMs by developing algorithms that leverage off-policy data to improve training efficiency and diversity, while preventing saturation in terms of plasticity and exploration.

My recent work includes co-leading the post-training of a fully open source 70B LLM (Apertus 70B, Project Apertus), leveraging diversity to improve test-time scaling and sampling in reinforcement learning (SR with the Gemma post-training team at Google DeepMind), deriving a calibrated offline & off-policy RL fine-tuning algorithm for LLMs (Quantile Reward Policy Optimization, NeurIPS 2025), and exposing the connection between plasticity, trust-region, and off-policy collapse (No Representation, No Trust, NeurIPS 2024).

I have developed several codebases from scratch with significant open-source contributions and infrastructure work such as a scalable code execution sandbox for QRPO and a scalable offline post-training pipeline for Apertus 70B. I am also proud of building infrastructure for reproducible ML (Python Machine Learning Research Template) and conducting award-winning reproducibility studies (ReScience 2023).

Background

I hold an MSc in Advanced Computer Science from the University of Oxford (Reuben College inaugural cohort 2021-2022). My MSc thesis on Multi-Agent RL and policy gradient methods was supervised by Mingfei Sun and Prof. Shimon Whiteson at the Whiteson Research Lab (WhiRL).

I validated a BSc in Mathematics and Computer Science from Ecole Polytechnique (inaugural cohort 2017-2020) and was a visiting student at EPFL, the University of Toronto, and Stanford University, where I focused on software engineering and entrepreneurship.

I was a PhD student researcher at Google DeepMind (Paris) with the Gemma post-training team hosted by Alexandre Ramé, working on leveraging diversity to improve test-time scaling and sampling in reinforcement learning. I also interned as an applied scientist at Quincus to optimise middle-mile logistics using RL, and as a software engineer at Amazon building AI tools for Amazon Transportation Services.

At EPFL, I was the president of CUBAliente (2024), the latin social dancing association. At Oxford, I led the Careers team of the Oxford Artificial Intelligence society (OxAI) (2022) and played as a libero for the men’s Blues volleyball team (2021-2022). At Ecole Polytechnique, I was elected in the first board of the L’ORE (2018), the Bachelor’s program student society.