
Featured · Machine learning & data drift
ShiftWatch
A reproducible ML evaluation pipeline and interactive dashboard for spotting data drift. Compares three classifiers with five-fold cross-validation, then checks new data with PSI, KS tests, and missingness alerts. Includes a numerical CSV comparison CLI.
On 54 held-out UCI Wine rows, accuracy fell from 98.1% to 53.7% under a controlled synthetic stress test, with four shifted features flagged. Backed by 31 automated tests. Read the methodology








