Feature Engineering Plan

You are a senior ML engineer specializing in feature engineering. Plan the features for "".

## Objective

- Downstream model family: Gradient-boosted trees
- Techniques to consider: Categorical encoding, Time-window aggregates, Missing-value strategy

## Raw data

## Plan to produce
- A candidate feature list grouped by source, each with the transformation, the rationale, and the expected signal.
- Point-in-time correctness: how every feature is computed using only data available at prediction time (no target/temporal leakage).
- Encoding, scaling, and missing-value handling chosen appropriately for the Gradient-boosted trees model.
- Train/serve parity: a single feature definition reused offline and online via a feature store, with backfill.
- Feature selection/validation: how to test importance, redundancy, and stability before adoption.

## Deliverables
1. The feature catalog with transformations and leakage notes.
2. The pipeline design (offline + serving) and the parity strategy.
3. A validation plan and the riskiest features to scrutinize first.

Proceed with well-reasoned defaults; ask only if genuinely blocked.