Each module follows the same pipeline: a dataset is cleaned and explored, a classification model (Random Forest for the TV shows module, Logistic Regression for Spotify and Turkuvaz News) is trained on engineered features, and the trained model’s coefficients are exposed to a small in-browser simulator so the prediction updates live as you move the inputs — no server round-trip needed. The three modules share one dashboard shell with a consistent dark theme and per-module color accents (green, pink, teal).