CV
Education, research experience, industry experience, and technical background.
Contact Information
| Name | Salah-Eddine Rizki |
| salaheddinerizki11@gmail.com |
Experience
-
2025 - 2025 Bordeaux, France
Data Scientist Intern · Recommendation Systems
Betclic Group
Investigated sequential recommendation and foundation-model approaches for large-scale interaction data.
- Reviewed recent recommendation-system literature and implemented a BERT4Rec-based modeling pipeline.
- Worked with more than 20 million interactions, 4,000 games, and 500,000 users.
- Improved offline ranking metrics by 2.5× and contributed to a 15% increase in engagement.
- Developed reproducible workflows with Docker, AWS SageMaker, Airflow, and Jenkins.
-
2024 - 2024 Paris, France
Machine Learning Engineer Intern
AEGLE
Designed and evaluated anomaly-detection methods on operational data.
- Trained and evaluated models on more than 50,000 records, reaching 94% accuracy.
- Reduced false positives by 15% and improved F1 score by 10% through feature and model analysis.
-
2023 - 2023 Casablanca, Morocco
Data Scientist Assistant Intern
OCP Group
Built analytical tools for monitoring industrial operations.
- Developed Python and SQL dashboards tracking 20 KPIs across five production units.
Summary
Education
-
2025 - 2029 Cergy, France
PhD
CY Cergy Paris University · ETIS Laboratory
Artificial Intelligence
- Research direction: explainable graph diffusion models.
- Advisor: Vassilis Christophides.
-
2024 - 2025 Cergy, France
Research Master's degree
CY Cergy Paris University
Data Science and Machine Learning
- GPA: 3.35/4.00
-
2022 - 2025 Cergy, France
Engineering degree
ENSEA
Signal Processing and Artificial Intelligence
- GPA: 3.55/4.00
Research
Publications
-
2026 Counterfactual Explanations for Equivariant Graph Neural Networks: CE-EGNN for Molecular Property Regression
Counterfactual explanation method for equivariant graph neural networks in molecular property regression.
Presentations
Projects
-
Message-Passing GNNs from Scratch
Transparent implementations of GCN and GAT in pure PyTorch, without PyTorch Geometric, with controlled classification, regression, and oversmoothing experiments.
- Sparse edge-index message passing
- Reproducible GCN/GAT comparisons
-
Financial Time-Series Transformer
Transformer-based modeling experiments for financial time series.
-
Fraud Detection
Machine-learning pipeline for fraud detection and model evaluation.
Skills
Research areas: Graph representation learning, geometric deep learning, graph generative models, explainable AI, molecular machine learning, MLIPs
Machine learning: PyTorch, deep learning, graph neural networks, transformers, recommendation systems, time-series modeling
Engineering: Python, SQL, Docker, AWS SageMaker, Airflow, Jenkins, Git