AI Research Engineer at INRIA specializing in Large Language Models (LLMs), reinforcement learning, and causal reasoning. Building on a background in applied mathematics, I co-develop open-source procedural data generation tools and supervise undergraduate research projects. Highly motivated to leverage this expertise for a PhD position focused on Agentic AI and Kantian evolutionary game theory.
Jan 2025 - Present
Lille, France
Jan 2025 - Present
Jan 2024 - Jul 2024
Le Blanc, France
Jan 2024 - Jul 2024
2022 - 2023
Milano, Italy
2022 - 2023
2022 - 2023
2019 - 2020
Nancy, France
2019 - 2020
2020 - 2020
Nancy, France
2020 - 2020
2020-2023 M.Sc. in Statistical LearningMention: 105/110 out of 110Taken Courses:
| ||
French graduate school of EngineeringMention: 3.6/4 out of 4Taken Courses:
| ||
2016-2018 Classes Préparatoires aux Grandes Écoles (Mathematics and Physics track) |
An open-source Python library for generating synthetic procedural data for LLM training (SFT + RLVR). Implements structural causal models supporting observational, interventional, and counterfactual queries based on Judea Pearl’s Ladder of Causation.
A live Streamlit playground for reasoning_core: pick one of its 40 procedural task families (logic, sets, graphs, code execution, planning…), choose a difficulty level, then try to solve the generated problem yourself and get scored instantly. Source on GitHub.
A production site for a working farm selling direct to its customers: product catalogue, pre-orders with a choice of collection point and day, stock and order management, customer ↔ farm messaging, printable picking sheets and sales statistics. Card payment handled through Stripe. Built with Node/Express, EJS and SQLite, and self-hosted on my own VPS behind Caddy.
Master thesis exploring the integration of Kant’s categorical imperative into evolutionary game theory. Developed a population dynamics model combining selfish gain with moral cooperation. Led to a published paper in the International Game Theory Review.
Three research projects conducted at Politecnico di Milano: Computer Vision for pulmonary disease detection, categorization of heart failure therapies, and a non-parametric statistical study on tobacco consumption trends across OECD countries.
Kant’s categorical imperative states: “Act only according to that maxim whereby you can at the same time will that it should become a universal law.” Game Theory has recently borrowed this idea from moral philosophy to introduce a new driver in strategic decisions — a motivation that adds cooperation to selfish reasoning. This paper presents a model describing the dynamics of a population whose strategies evolve not only from selfish gain, as in the replicator dynamics, but also from a moral inclination toward cooperation. We apply our model to well-known game-theoretic problems.
