The Data Scientist / ML Engineer is the management-side builder of datasets and models. Role 1.1 Directs purpose and appetite. Role 1.4 owns the risk file. Role 2.2 runs the model in production. This role turns a signed purpose and requirements pack into lineage-documented data, reproducible training, evaluation against that purpose, and validation metrics Operate can actually run.

Standing duties

• Implement models and datasets only against the last signed intended purpose (ACT02). Stop and raise Act when the work would change purpose in fact.
• Produce lineage, representativeness, bias, robustness and experiment records that EST14, EST12 and the Annex IV file can cite.
• Design continuous-validation metrics and guardrails at build time. Role 2.2 operates them; this role must not hand over a model with no reopen rule.
• Refuse to treat a model card, fairness score or notebook as residual-risk acceptance or as a control.
• Capture experiment knowledge so Role 17.1 and the QMS can reuse it — not as personal folklore.

Accountable to: AI Governance Lead / Head of AI Development (or GM: Data Science)

Primary domain: Implement (model and dataset build). Designs validation metrics Operate will run. Does not Direct or accept residual risk.

Linked EN 18286 clauses: 4.3, 4.4.2, 8.4.2, 8.5, 8.8

Format: 3 hours of training, EUR 200.