Machine Learning • Quantitative Finance
🎓 M.A. in Statistics at Columbia University (4.19/4.33)
🎓 B.S. in Applied Mathematics & Statistics from Stony Brook University (3.99/4.0)
| Project | Summary |
|---|---|
self-instruct-gpt4mini |
Current working project |
xrt-trading-case-study |
Built and backtested deviation- and volatility-based trading strategies on XRT, analyzing hit rates, Sharpe ratios, and robustness, concluding no persistent alpha after risk adjustments. |
mnist-image-classification |
Comparing Lasso, Naive Bayes, Ridge, SVM, and Group Lasso |
tcga-brca-analysis |
Unsupervised analysis of TCGA Breast Cancer (BRCA) gene-expression data |
trees-ensembles-neural-networks |
Decision trees, boosting, and neural networks on UCI income data with tuning, overfitting analysis, and feature interpretation. |
algorithmic-trading |
Systematic trading pipelines using ML and time-series cross-validation |
latent-semantic-clustering |
UMAP + EM-GMM clustering of book chapters via NLP frequency vectors |
quantitative-finance |
BSM & Heston option pricing, Monte Carlo simulations, VaR, etc |
crime-predictor-analysis |
Predicting crime using UCI community features; LASSO, Ridge, Elastic Net, kernel regression + manually implemented CV |
sepsis-prediction |
Cleaned & merged using SQL, then applied ML pipeline to CUMC + NYP secure patient-level dataset; HIPAA-compliant experiments using Azure Secure Environment; certified. |
Supervised Learning & Statistical Modeling: LASSO, Ridge, Elastic Net, Logistic Regression, LDA, ARIMA, Group Lasso, etc
Dimensionality Reduction & Feature Analysis: PCA, UMAP, t-SNE, Spectral Embedding, MDS, NMF, Kernel PCA
Unsupervised Learning & Clustering: KMeans++, Gaussian Mixture Models (GMM), Spectral Clustering, Hierarchical Clustering
Languages: Python, SQL, HTML, R, MATLAB
Libraries: PyTorch, TensorFlow, scikit-learn, XGBoost, Numpy, Pandas, Statsmodels
Visualization: Plotly, Matplotlib, Seaborn, Streamlit
Workflow: VSCode, Git/GitHub, Google Colab (for GPU compute), LaTeX
Infra: Azure, APIs, GitHub Actions
🌐 Connect With Me
LinkedIn
GitHub
[email protected]
"Averaged over all possible data-generating distributions, every classification algorithm has the same error rate."
— David H. Wolpert, No Free Lunch Theorems for Optimization