Mohamed Rahmouni

Senior Ml Engineer @BIOSerenity

Paris, FR

Led development of BioSerenity-E1, a self-supervised Transformer-based foundation model for EEG — architecture design, pretraining strategy, fine-tuning, and evaluation framework- Architected distributed training infrastructure: PyTorch DDP across multi-GPU nodes, Slurm-based HPC job orchestration, data sharding and async prefetching for throughput optimization- Built end-to-end data pipelines (Ray, WebDataset) processing millions of time-series records, designed for reproducibility and fast iteration across experiments- Set up MLOps tooling: experiment tracking (MLflow), model versioning, CI/CD (GitLab), containerized training environments (Docker, AWS SageMaker)Demonstrated +17% AUPRC improvement in low-data regimes (<10% labeled data), validating SSL pretraining as a practical strategy for label-scarce clinical settings- Co-authored 2 publications (arXiv) on foundation model performance for clinical EEG screeningMentored junior data scientists and engineers — code review, experiment design, debugging training runs

Pioneer school of ELkef

Baccalaureate, Mathematics

INSAT - Institut National des Sciences Appliquées et de Technologie

Engineer’s Degree, Information Technology