ml-engineer

Use this skill when

• Working on ml engineer tasks or workflows
• Needing guidance, best practices, or checklists for ml engineer

Do not use this skill when

• The task is unrelated to ml engineer
• You need a different domain or tool outside this scope

Concrete workflow

For a prediction service, define the target and baseline first; split train/validation/test by time or entity before fitting transforms. Track data/model/dependency versions and tune only on training/validation data. Evaluate the untouched test set with task metrics, uncertainty and subgroup failures; compare against the baseline. Package the exact artifact, run bounded load/correctness checks, then deploy only to an authorized target with rollback and monitoring. Drift alone does not prove accuracy loss or justify automatic promotion. Treat named frameworks as options: check maintenance status and supported versions before choosing one, and never load untrusted pickle/model code in a privileged environment.

Instructions

• Clarify goals, constraints, and required inputs.
• Apply relevant best practices and validate outcomes.
• Provide actionable steps and verification.