Open to opportunities

yaakov levin

@yaakovlevin

I’m a self-taught data analyst building fraud detection and forecasting pipelines with Python, SQL, and business-ready dashboards.

Israel

Message

What I'm looking for

I’m looking for analyst roles in fintech, risk, or enterprise analytics where I can build end-to-end pipelines, predictive models, and business-ready dashboards—especially in fraud detection and time-series forecasting—using Python, SQL, and rigorous statistical methods.

I’m a self-taught data analyst with hands-on experience building end-to-end analytical pipelines, predictive models, and business-ready dashboards. I specialize in fraud detection, financial time-series forecasting, and turning large, complex datasets into actionable insights. I combine rigorous statistical thinking with practical Python and SQL to deliver measurable business value.

In exchange-rate forecasting, I built SARIMAX and HistGradientBoosting models using macro drivers (DXY, NDX, VIX, crude oil, and interest rate differentials). I engineered a geopolitical conflict dummy variable (Oct 7, 2023 onset) as a statistically significant predictor, while applying strict lag-based leakage prevention and resolving data contamination from EUR cross-pairs.

For retail forecasting, I designed a two-stage pipeline with a DuckDB SQL cleaning layer and modeling in MixedLM and SARIMAX, reaching ~3.5% MAE across 454 cities. For credit card fraud, I trained an XGBoost classifier with GridSearchCV on ~8.5M rows, reached 94% recall on the fraud class, added SHAP explainability, and deployed a Flask + React app to output real-time fraud probabilities and tiered risk labels.

Experience

Work history, roles, and key accomplishments

USD/ILS Forecasting Analyst

Portfolio Projects

Jan 2024 - Jan 2025 (1 year)

Built SARIMAX and HistGradientBoosting models to forecast USD/ILS using macro drivers; engineered a geopolitical conflict dummy (Oct 7, 2023) that was statistically significant and applied strict lag-based leakage prevention. Resolved cross-pair contamination issues and delivered a currency-risk forecasting tool for fintech and Israeli importers/exporters.

XGBoost Feature Engineering SHAP Explainability Python (Pandas NumPy SQL

Credit Card Fraud Detection

Portfolio Projects

Jan 2023 - Jan 2024 (1 year)

Trained an XGBoost fraud detection model with GridSearchCV on ~8.5M synthetic transactions, achieving 94% recall on the minority fraud class. Added SHAP explainability and Haversine geospatial features, and built a Flask + React app for real-time fraud probability scoring with an ROI framing of ~$11.9M net savings.

XGBoost Fraud Detection Model Tuning (GridSearchCV)SHAP Explainability Geospatial Feature Engineering (Haversine)Flask React Pandas

Iowa Liquor Sales Analytics

Portfolio Projects

Jan 2024 - Present (2 years 5 months)

Created a DuckDB SQL cleaning layer feeding MixedLM and seasonal SARIMAX models for Iowa liquor sales, achieving ~3.5% MAE across 454 cities. Produced safety-stock recommendations using a 95th-percentile z-score approach and published Tableau, Kaggle, and GitHub deliverables.

DuckDB Exogenous Variables EDA Python (Pandas NumPy Matplotlib