I work on machine learning, reinforcement learning, and natural language processing. Interested in out-of-distribution generalization, information theory, and the emergence of structure in complex systems. Further insights can be found on my blog The Prior Problem.
Beyond research, I aim to design intelligent systems that assist people meaningfully and contribute positively to society.
Feel free to reach out at: delineauj [at] gmail [dot] com
- B.Sc. in Pure Mathematics, EPFL - Probability & Measure Theory
- M.Sc. in Applied Mathematics, EPFL - Machine Learning Theory
Implemented key reinforcement learning algorithms: PPO, VFE, and SAC, from scratch and benchmarked them across multiple environments.
Personal Project · May–June 2025
Rebuilt GPT-2 architecture from the ground up to deeply understand transformers, and trained it on a poetry dataset.
Personal Project · March–April 2025
Predicted monthly electricity demand for thousands of consumers using LightGBM and advanced feature engineering.
ETH Hackathon · 4th/11 · April 2025
Developed a mathematical and perceptual framework for ARC-AGI. Proposing a classification-based abstraction approach inspired by human reasoning.
EPFL Semester Project (10 ECTS) · Sep 2024–Jan 2025