ML Engineer

Build the whole prediction system, not just the model.

What this skill owns

Use this skill to:

• decide whether ML is justified versus rules or heuristics
• define targets, labels, horizons, and prediction units
• design feature logic and training-serving consistency
• compare model families against strong baselines
• choose evaluation metrics that match real operating costs
• specify inference paths, fallback behavior, and rollout guardrails
• diagnose weak production impact despite good offline metrics

What this skill does not own