🎓 CS & Applied Math @ NYU • AI/ML
Probability of default modeling on 1.3M+ loans with production decision logic.
- Built PD model achieving 0.72 ROC-AUC with leakage-resistant, out-of-time validation
- Engineered 50+ features: DTI, utilization bins, loan-to-income ratios, interaction terms
- Mapped predicted probabilities to credit decision tiers (Approve / Manual Review / Reject)
- Identified sub-grade and interest rate as dominant risk factors; addressed 20% class imbalance
Methods: Logistic Regression, Random Forest, Gradient Boosting, class weighting
Quantitative risk metrics and sentiment-driven trading signals.
- Calculated Value at Risk (Historical, Parametric, Monte Carlo) across 7 equity positions
- Built GARCH(1,1) volatility forecasts for forward-looking risk estimates
- Developed sentiment signal pipeline: FinBERT on financial news → z-score normalization → daily aggregation
- Trained XGBoost direction classifier with 100+ features using walk-forward validation to prevent lookahead bias
Methods: VaR, GARCH, NLP sentiment analysis, time series forecasting
Business intelligence platform with dimensional modeling and KPI automation.
- Modeled customer LTV, cohort retention, and funnel conversion using star schema design
- Implemented SCD Type 2 dimensions for point-in-time historical accuracy
- Reduced reporting cycle time with automated KPI pipelines; benchmarked at 50M rows with 8-40x query speedup
Methods: Dimensional modeling, cohort analysis, ETL orchestration
CI/CD release gate for LLM applications: quality and compliance checks before production.
- Built evaluation framework with PII detection (credit card Luhn validation, SSN format checks)
- Multi-layer prompt injection detection for model security
- 159 unit tests, multi-provider support (OpenAI, Anthropic, Azure)
| Category | Tools |
|---|---|
| Languages | Python, SQL, Java, C++, TypeScript |
| Data & ML | pandas, NumPy, scikit-learn, XGBoost, Spark, Kafka |
| Statistics | Regression, classification, hypothesis testing, time series, probability |
| Visualization | Matplotlib, Seaborn, Plotly, Tableau, Power BI, Streamlit |
| Data Infrastructure | PostgreSQL, AWS (S3, Redshift), dbt, Prefect, Airflow, Docker |
| Quantitative Finance | VaR, GARCH, PD modeling, credit risk, volatility forecasting, Monte Carlo |
💡 I like building things that actually work, then figuring out why they work.