Detecting Data Drift (PSI & KS)

Day 73 of 100 Days of MLOps. Ground truth is slow, but the first sign of decay often shows up in the model's inputs — the live data starts to look different from the training data, long before you know if predictions were right. Today you measure that with two standard tools: the Population Stability Index (PSI) and the Kolmogorov–Smirnov test. You'll compare a stable sample and a drifted one and watch both methods agree — an early-warning system that needs no labels. Runs 100% locally on any OS.