nancy boukamel

• Build end-to-end data pipelines with Pandas and custom imputation strategies to handle missing sensor data at scale. • Developed transformer and LSTM-based forecasting models across 10 years of energy consumption data (>8000 profiles), achieving MAPE <2%. • Accelerated hyperparameter tuning by 35% using Optuna and Scipy optimizers. • Contributed to code reviews, cleaned code standards, and interbal technical documentation.

• Refined SQL pipelines for large financial datasets (100s of millions of rows), reducing execution times by ~40%. • Built a customer segmentation model using XGBoost (ROC-AUC = 0.85) to support regional risk scoring. • Used PCA and t-SNE to compress high-dimensional data and visual clusters, validated by silhouette scores. • Led a Proof of Concept using embedding models (e.g. Word2vec, Sentence-BERT) for semantic-similarity and retrieval across financial text data.

Engineered and deployed full-stack cloud solutions leveraging AWS (Step Functions, IAM, Cognito, CloudWatch), Docker, and AngularJS/NoSQL. Designed robust CI/CD pipelines using Jenkins and AWS CodePipeline, which cut deployment time by 50%. Built real-time orchestration workflows for financial transactions, reducing system downtime by 25%. Improved data refresh latency by ~30% via dynamic dashboards. Containerized services ensured consistent deployment. Collaborated in Agile Scrum teams using Git.