CV

Education

M.Sc. Mathematics, Vision and Learning (MVA)

École normale supérieure Paris-Saclay · Gif-sur-Yvette, France · 2025-2026

• Relevant coursework: optimal transport, convex optimization, computational statistics, geometric data analysis, graph machine learning, random matrix theory, inverse problems, generative models, stochastic calculus and robotics.

Diplôme d’ingénieur (M.Sc. equivalent), Mathematics and Computer Science

Mines Paris - PSL · Paris, France · 2022-2026

• Relevant coursework: measure theory, probability, statistics, machine learning, generative models, functional analysis and signal processing.

CPGE MPSI-MP* - intensive mathematics and physics program

Lycée Hoche · Versailles, France · 2020-2022

• TIPE: Radon transform for tomography.

Research Experience

Research Intern - Weak error analysis for discrete diffusion samplers

École normale supérieure - Department of Mathematics and Applications (DMA) · Paris, France · 2026-present

Supervisors: Julie Delon, Rémi Gribonval and Gabriel Peyré.

• Derived the leading weak-error term of the matrix Euler discretization for Bayesian time reversals of finite-state continuous-time Markov chains.
• Obtained a spectral representation showing how the bias depends on the corruption generator, noise schedule, data distribution, and test observable.
• Analyzed the interaction between terminal mismatch, reverse-rate perturbations, and Euler bias on a two-state graph.

Research Intern - PAC-Bayesian generalization bounds

Inria · Lyon, France / London, UK · 2025 (4 months)

Supervisors: Antoine Gonon, Rémi Gribonval, and Benjamin Guedj.

• Studied neuron-wise rescaling symmetries of ReLU networks and their effect on PAC-Bayes complexity terms.
• Formulated PAC-Bayes bounds in an invariant lifted representation and analyzed their guarantees through data processing.
• Implemented KL-based optimization procedures and evaluated them on neural-network experiments using PyTorch Lightning and Weights & Biases.

Research Intern - Programmable origami metamaterials

Harvard SEAS - Bertoldi Group · Boston, USA · 2024 (5 months)

• Modeled compatibility constraints for programmable origami patterns.
• Studied multistable transitions using Abaqus simulations and experimental validation on macro- and microscale prototypes.
• Implemented a Python pipeline generating DXF fabrication files from geometric design parameters.

Preprint

• D. Rouchouse, A. Gonon, R. Gribonval, and B. Guedj. (2025). “Non-Vacuous Generalization Bounds: Can Rescaling Invariances Help?”. arXiv. arXiv:2509.26149. Read the paper.

Additional Experience

Computer Vision Intern

Scortex - TRIGO Group · Paris, France · 2024-2025 (6 months)

• Adapted diffusion- and distillation-based computer vision methods for high-speed industrial anomaly detection.
• Benchmarked and deployed selected models under real-time constraints using PyTorch and MLflow.

Oral Examiner in Mathematics

French Ministry of Education · Versailles, France · 2023-2024 (6 months)

• Prepared and assessed oral mathematics examinations for students preparing engineering-school entrance examinations.

Technical Skills

• Programming: Python (PyTorch, PyTorch Lightning, NumPy, scikit-learn), working knowledge of C/C++.
• Tools: Git, Linux, Docker, LaTeX, MLflow, Weights & Biases.
• Languages: French (native), English (TOEFL iBT: 110/120).