I am currently pursuing a Ph.D at the LITIS Lab, INSA Rouen, with a research focus on time series anomaly detection, forecasting and the development of foundation models for zero-shot forecasting. I have a strong background in Mathematics, with a Bachelor’s degree in Mathematics from Paris-Saclay University and a Master’s degree in Mathematics and Artificial Intelligence from the same institution.

News ⭐️

• 10-2025 - Completed the Deep Learning Optimisé sur Jean Zay program, focusing on cutting-edge techniques for optimizing deep learning models on HPC infrastructure using PyTorch. Gained hands-on experience and knowledge with multi-GPU and multi-node training, from standard Distributed Data Parallel (DDP) to advanced 4-D parallelism, as well as mixed-precision training, large-batch strategies, communication optimization, and performance profiling.

• 10-2025 - Participated in the Fête de la Science de Rouen 2025 where I developed an interactive game that challenges players to predict future signal values and compete against a zero-shot forecasting model I designed: PatchFM. The objective is simple — beat the model by making more accurate predictions. The game code is available here.

• 09-2025 - Participated in the EUSIPCO 2025 conference presenting my paper PatchTrAD in a 15-minute talk. The slides can be found here.

• 07-2025 - Participated in the Hi! Paris Summer School 2025 presenting my paper PatchTrAD in a poster session. The poster can be found here.

• 05-2025 - PatchTrAD has been accepted at EUSIPCO 2025.

Me, Myself & Maths 🔢

I am passionate about Artificial Intelligence. In my spare time, I enjoy building state-of-the-art Machine Learning models and exploring the mathematical theories behind them. One of my favorite architectures is the Variational Autoencoder, introduced in the paper Auto-Encoding Variational Bayes. I’m particularly drawn to VAEs because they elegantly combine Bayesian inference with deep learning. Working on a project involving this architecture was a turning point, it sparked my desire to pursue a Ph.D, further reinforced by the professional experience I gained during my internships. In addition, I enjoy sharing my knowledge and passion for mathematics by working as a tutor for students.

Favourite Papers 📝

1. Auto-Encoding Variational Bayes
   Obviously, it holds the first position.

2. Neural Tangent Kernel: Convergence and Generalization in Neural Networks
   Infinitely wide neural networks behave like kernel regression, and their training dynamics can be analyzed through their corresponding kernels. I appreciate this paper because it provides a solid theoretical foundation for wide neural networks, supported by extensive mathematics.

3. PatchTST: A Time Series is Worth 64 Words: Long-term Forecasting with Transformers
   This paper proposes a powerful Transformer architecture for time series forecasting by dividing sequences into patches, analogous to tokens in LLMs. Following its publication, the patch-based paradigm gained significant attention in time series modeling. This approach enables training time series models with architectures closely aligned to those used in LLMs.

4. Denoising Diffusion Probabilistic Models
   I really appreciate this paper; the Bayesian inference theory underpinning diffusion models is both fascinating and enjoyable to understand.

5. Regularized Evolution for Image Classifier Architecture Search
   I was torn between this paper and the original Transformer paper. This paper employs a simple evolutionary algorithm to discover highly effective architectures for image classification. The underlying intuition is straightforward, yet it proves to be an incredibly powerful strategy.

Sports 🚴🏼

The main sport I pursued as a teenager was cycling, and I participated in numerous national races during my teenage years. I won several department and regional races, including being crowned Department Champion once and Regional Team Champion once. When I stopped cycling, I transitioned to streetlifting, a strength sport focused on muscle-ups, pull-ups, dips, and squats. I have competed twice in the French Dips Championship and I have been the First Step French Champion of Streetlifting in 2024 in the -80 kg category. I have also worked as a trainer, helping athletes reach higher levels by providing them with tailored programs and weekly progress tracking.