Jad Yehya

About

New look coming soon!

I am a second-year PhD candidate in the MIND team at Inria Saclay, advised by Thomas Moreau and Matthieu Kowalski. I work on machine learning and deep learning for time series, including foundation models, anomaly detection, and forecasting. In parallel, I am completing an MBA at Collège des Ingénieurs.

Before my PhD, I studied computer science and applied mathematics at Sorbonne University, where I earned a bachelor's and a master's degree. I also studied in Milan during my master's.

jad.yehya@inria.fr Google Scholar GitHub LinkedIn

Research interests

Time series, foundation models, tokenization, anomaly detection, forecasting.

Selected Publications

RoseCDL: Robust and Scalable Convolutional Dictionary Learning for Rare Event and Anomaly Detection

Jad Yehya*, Mansour Benbakoura*, Cédric Allain, Benoît Malezieux, Matthieu Kowalski, and Thomas Moreau.

29th International Conference on Artificial Intelligence and Statistics (AISTATS), 2026.

Paper · Code

Workshop papers

RoseCDL 1: Robust and Scalable Convolutional Dictionary Learning for Rare Event Detection

Jad Yehya, Mansour Benbakoura, Cédric Allain, Benoît Malezieux, Matthieu Kowalski, and Thomas Moreau.

NeurIPS 2025 Workshop on Learning from Time-Series for Health (TS4H), spotlight.

Software and open source

RoseCDL. Robust and scalable convolutional dictionary learning.
TSAD Benchmark. A benchmark for time-series anomaly detection methods.

Open-source contributions

TSFM Benchmark. I contribute to a benchmark for time-series foundation models across different tasks and datasets.

Talks

Tackling Time Series Anomalies: Unsupervised Detection and Evaluation Strategies

IRFM - CEA, April 2025.

Other activities

I organize a weekly meeting at my lab where PhD students discuss new tools, technology, and AI agents. Outside research, I play piano, compose music, and take (too many) photos.