#Machine Learning
Understanding Calibration Curves: Why They Matter and How to Read Them
What calibration curves show, how to read deviations from the diagonal, and how to fix a miscalibrated classifier.
September 22, 20265 min read
Honeywell AOG Zero: Building an APU Health Monitoring System in a Week
A hackathon case study on building a real-time aircraft APU health dashboard with Next.js, FastAPI, and PostgreSQL — and the trained ML model that never made it to production.
September 14, 20266 min read
Forecasting Cash Flow with Uncertainty, Not Just Averages
Separate income/expense LightGBM regressors, time-aware features, walk-forward validation, and 80% confidence intervals make cash-flow projections actionable.
September 11, 20265 min read
Transaction Categorization That Survives Cold Starts
A hybrid categorization system with rule-based fallback, LightGBM classifier, and confidence-gated ML predictions—ensuring sensible behavior from day one.
September 10, 20265 min read