Blog

A Python ML framework built on monadic design: DataPod state carriers, composable trained transformers, and reproducible pipelines from research to production.

Markov Chains turn binary default labels into continuous default probabilities, improving risk models on small data and enabling fairer portfolio comparisons.

Hybrid CNN and LSTM models beat ARIMA, RSI, and MACD on WTI crude: LSTM-CNN reached 17.8% average annual ROI with the best Sharpe ratio we tested.

How I replaced manual Excel-to-simulator data transfer with a Python tree-modeling tool for offshore pipeline design, cutting $49.8M from the investment plan.

A linear programming optimization app for a gas field of nearly 1,000 wells: the LP model, the data pipelines behind it, and the 15.2% production lift.