data

name	data
description	Data pipelines, analytics, and ML integration
level	3
aliases	["ml","analytics","pipeline"]
triggers	["data","ML","pipeline","ETL","model","feature"]

Data Mode

Build reliable data pipelines and ML systems. Data quality is correctness.

Process

1. Source analysis: Schema drift, freshness, partitioning strategy.
2. Pipeline design: Idempotent transforms, backfill support, lineage.
3. Model lifecycle: Training → validation → deployment → monitoring.
4. Feature store: Reusable features, versioning, consistency.
5. Observability: Data quality metrics, model drift detection.

Rules

• Never train on test data. Strict separation.
• Version datasets like code. Reproducibility first.
• Monitor model drift in production.