About Me

I’m Thomas Massena, a PhD student at IRIT (Institut de Recherche en Informatique de Toulouse) and SNCF.

Before that, I was a naïve intern at IRT Saint Exupéry who thought he would be designing guitar pedals for a living. My research experience alongside the endlessly curious Louis Béthune was enough to make me completely fall in love with the world of research.

Nowadays, my research mainly focuses on trustworthy and certifiably robust methods for machine learning, with a particular emphasis on computer vision tasks. I work on developing neural networks with provable guarantees, including Lipschitz-constrained architectures for differential privacy and robust conformal prediction.

I have recently started working on optimization methods for large neural networks, developing several interesting theories and improvements to the Muon optimizer along the way.

Research Interests

• Optimization for deep learning
• Certifiable robustness in deep learning
• Differential privacy for neural networks
• Conformal prediction and uncertainty quantification
• Lipschitz-constrained neural networks
• High-performance parallel implementations

Community

I have served as a reviewer for CVPR (2025, 2026), ICML (2025), and NeurIPS (2026), among others.

I supervised Yohan Le Morhedec (Centrale Supelec) during an internship at SNCF.

Contact

Feel free to reach out via email at thomasmassena@gmail.com.

Miscellaneous

I recently started birdwatching and will try to keep my bird life list reasonably up to date.