Feature Engineering
Crafting better input features from raw data — often the highest-leverage work in classical ML.
Feature engineering is transforming raw data into input features that make patterns easier for a model to learn — deriving ratios, aggregations, date parts, encodings, and interactions. In classical ML it is often the single highest-leverage activity: a good feature can beat a fancier model.
Worked example: from a raw timestamp you engineer “hour of day,” “is weekend,” and “days since last purchase” — features that expose the signal a model could not easily extract from the raw epoch number. Gotcha: feature engineering is where data leakage sneaks in — accidentally using information not available at prediction time (a feature computed from the future, or from the label) inflates offline scores and collapses in production; and deep learning reduces but does not eliminate the need for it, especially on tabular data.