26.06.09 New project: a statistical arbitrage module that screens equity pairs with Engle-Granger and Johansen cointegration tests, models the spread as an Ornstein-Uhlenbeck process, replaces the static OLS hedge ratio with a Kalman filter dynamic estimate, and backtests z-score entry/exit signals with transaction costs and cointegration stability monitoring — with a Typer CLI and Streamlit dashboard. Visit the project page.

26.06.09 New project: a WGAN-GP trained on cross-asset daily log returns to learn the empirical joint distribution without parametric assumptions — benchmarked against historical simulation and multivariate normal bootstrapping across tail diagnostics, correlation structure, GARCH volatility persistence, PCA alignment, and out-of-sample VaR coverage — with a Typer CLI and Streamlit dashboard. Visit the project page.

26.06.08 New project: an XGBoost and LightGBM ensemble that predicts next-month cross-sectional stock returns from 18 fundamental, momentum, and liquidity factors — with walk-forward validation, SHAP attribution, decile portfolio construction, IC/ICIR/Sharpe evaluation, and signal decay analysis — with a Typer CLI and Streamlit dashboard. Visit the project page.

26.06.07 New project: a Python module for yield curve analysis — bootstrapping spot rates from bond prices, interpolating curves (linear, cubic spline, and Nelson-Siegel), computing duration and convexity, and exploring four synthetic curve presets through a Streamlit dashboard. Visit the project page.

26.06.07 New project: a buy-side factor research toolkit that pulls SEC EDGAR XBRL financial statements, engineers cross-sectional equity signals (earnings yield, accruals, ROA, momentum, and more), ranks stocks into quantiles, and evaluates long-short spread performance via rolling IC backtests — with a Typer CLI and Streamlit dashboard. Visit the project page.

26.06.07 New project: a Python module implementing Black-Scholes and CRR binomial tree pricing for European and American options, analytical and finite-difference Greeks, implied volatility surfaces, multi-leg scenario analysis (call/put spreads, straddle, strangle), a Typer CLI, and a Streamlit dashboard. Visit the project page.

26.06.04 New blog post: a minimal compound interest calculator written in Rust, covering basic I/O, type parsing, and the standard compounding formula. Read it here.

26.06.02 New blog post on the dual nature of AI — how the same technology can act as both a tool that amplifies human judgment and an autonomous agent that supplants it — is now live here.

25.03.15 New project: a Python module that fits EWMA and GARCH(1,1) to daily asset returns, forecasts conditional volatility across 1-, 5-, 10-, and 21-day horizons, benchmarks each model against realized vol proxies with MSE/MAE/QLIKE, and classifies market periods as calm, normal, or stressed — with a Typer CLI and Streamlit dashboard. Visit the project page.

24.01.26 New project: a Python CLI portfolio risk engine computing rolling volatility, historical and parametric VaR/CVaR, stress tests (calibrated to 2008, 2022), Kupiec POF backtesting, and Brinson-style factor attribution across multi-asset portfolios. Visit the project page.

21.01.18 My research on Long/Short Global Macro Strategies — examining the construction and backtesting of a systematic long/short portfolio across global macro factors and asset classes — was just released. For more details, including the full report and methodology, visit the project page.

20.12.30 My research on predicting U.S. interest rate movements from Federal Reserve communications using natural language processing — combining topic modeling, sentiment analysis, and time series methods — was just released. For more details, visit the project page.

20.12.29 My research on Genetic Neural Networks — exploring the use of evolutionary algorithms to train neural network weights for portfolio management signals — was just released. For more details, visit the project page.