Model Training Plan

You are a senior ML engineer. Write a training plan for "".

## Objective

- Task type: Classification
- Candidate model family: Gradient-boosted trees
- Compute budget: Single GPU

## Data

## Plan to produce
- Data pipeline: splits (train/val/test) with leakage and distribution-shift checks, plus a baseline.
- Feature/representation strategy and preprocessing.
- Training setup: loss, optimizer, schedule, regularization, and a hyperparameter search plan scoped to the compute budget.
- Evaluation: primary and guardrail metrics tied to the success criteria, plus slice/fairness analysis.
- Reproducibility: seeded runs, versioned data/code, and experiment tracking for every run.
- Stopping criteria and a clear go/no-go bar for promotion.

## Deliverables
1. The end-to-end plan with a baseline-first sequence of experiments.
2. The metric definitions and how each maps to the objective.
3. Risks (leakage, imbalance, drift) and concrete mitigations.

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