Nancy J.

• Designed automated ETL pipelines in Python and SQL to integrate data from web analytics, transactions, and advertising platforms, improving data freshness and enabling near real-time marketing analytics • Developed a hybrid forecasting framework (XGBoost + multi-input ANN) to predict SKU-level demand, incorporating time-aware features such as promotion cadence and seasonality; reduced WMAPE by 5.6% and improved inventory accuracy • Implemented CI/CD pipelines using Airflow and MLflow to automate model retraining, tracking, and deployment • Implemented A/B testing and Diff-in-Diff analysis to quantify pricing elasticity and promotion effectiveness; revealed causal relationships between markdown strategies and customer conversion patterns • Communicated model insights and data narratives to cross-functional stakeholders through tailored visualizations (e.g. Power BI) and executive briefings, directly informing prioritization decisions on assortment strategy and vendor performance

• Engineered automated analytics pipelines in Python and SQL to consolidate user engagement and product metrics from multiple sources; improved data accessibility and enabled scalable insight delivery for campaign and feature performance tracking • Built modular Looker dashboards using BigQuery materialized views to monitor KPIs across funnels and cohorts; standardized metrics definitions to align product and marketing teams on shared performance objectives • Designed and analyzed multi-variant A/B tests, applying CUPED and propensity score matching (PSM) to control for variance and selection bias; produced causal insights that guided campaign optimization and UI rollout decisions

• Designed forecasting frameworks (ARIMA, Prophet) to project campaign ROI and customer retention trends; supported quarterly planning and marketing budget allocation with quantitative insights • Developed SQL-driven Tableau dashboards with modular query layers to track conversions, retention, and lifetime value across channels; improved cross-team visibility and reporting efficiency by 37% • Conducted market basket and churn analysis to identify high-value customer segments and optimize loyalty programs

• Led an end-to-end fraud detection pipeline: profiled demographics/income/education/transaction patterns, engineered and reduced features (correlation + PCA), addressed class imbalance with SMOTE, and trained KNN/Random Forest models—improving detection efficiency by 30% with higher recall on minority fraud cases • Delivered actionable insights and visual analytics to client stakeholders, supporting risk mitigation and strengthening compliance confidence.