Kaoutar Boulif

• Improved ROI by 10% by optimizing RTB pipeline for conversion probability prediction using Bayesian logistic regression. • Improved pipeline performance with feature engineering, memory optimization, and parallelized training, achieving 20% lower training cost. • Conducted R&D benchmarks comparing multi-conversion vs. binary models (Poisson, Gradient Boosting, Logistic Regression), testing advanced feature encoding and loss functions (MSE, SMAPE). • Collaborated cross-functionally with product and DevOps teams to productionize models using Flask APIs and Airflow DAGS.

• Processed and analyzed large-scale mortality datasets (1,400+ configurations) to build ARIMA forecasting models for crisis-based trend estimation. • Generated 2,800+ statistical parameters for actuarial risk assessment and presented findings at an international longevity conference. • Developed automated data cleaning and visualization dashboards using R, SQL and Tableau.

• Performed dimensionality reduction (PCA) and unsupervised learning (K-Means, Isolation Forest, IQR) on 51-feature security datasets. • Enhanced Fl-score by 4% through hyperparameter tuning and feature scaling. •Documented model reproducibility workflows and assisted researchers in publishing findings.