Machine learning has moved from experiment to infrastructure, and with that shift comes accountability. Regulators, auditors and customers now expect organisations to explain how automated decisions are made, show that data was lawfully obtained and demonstrate that models are tested, monitored and reversible. Teams that cannot do so risk fines, loss of trust and costly rework.

The Governance and Compliance in Data Science & Machine Learning in Practice Training Course joins the technical and control perspectives that are too often kept apart. Early sessions establish the governance operating model, policies, roles and the model lifecycle, together with the data protection principles that apply to training data. Participants then build models in Python and apply controls as they go, covering data quality checks, leakage prevention, validation design, fairness metrics, interpretability and stress testing.

The later days address production. Delegates learn how to document, approve, deploy and monitor models, manage drift and incidents, and prepare for audits. Each participant develops a full governance pack on a model of their choice, receiving feedback from the facilitator and peers, so that the course ends with working artefacts rather than notes.

By the end of the course, participants will be able to:

• Design a governance framework for data science and machine learning, with policies, roles, escalation routes and a risk-tiered model inventory.
• Apply data protection and privacy principles to the collection and use of training data.
• Test models for bias and fairness using appropriate metrics and mitigation methods.
• Explain model behaviour using SHAP, LIME and partial dependence techniques.
• Perform independent validation, including back-testing, sensitivity and stress tests.
• Implement reproducibility, versioning and lineage with tools such as Git, DVC and MLflow.
• Monitor deployed models for drift, performance decay and incidents.
• Prepare documentation and evidence packs for internal audit and regulators.

The programme is designed for people who build, approve or assure analytical models, including:

• Data scientists and machine learning engineers
• Model risk managers and independent validators
• Compliance, legal and data protection officers
• Internal and external auditors of analytics and technology
• Heads of analytics, data and digital transformation
• Product owners deploying automated decision systems
• Risk officers in banks, insurers, telecoms and health organisations
• Public sector data and policy teams using algorithmic tools

Participants finish the course with:

• A governance policy outline and model lifecycle for their organisation.
• A model inventory and risk tiering template.
• A completed model card, data sheet and validation report from their own project.
• A fairness and explainability testing notebook in Python.
• A monitoring plan with drift thresholds and incident response steps.
• An audit evidence checklist covering data, model and deployment controls.
• A data protection impact assessment template adapted for machine learning projects.
• A change control and re-validation checklist for models moving through MLflow environments.

The course combines governance workshops with technical labs, including:

• Python notebooks that apply a control to each stage of the model lifecycle
• Governance clinics where participants draft real policies and templates
• Peer validation exercises in which each person reviews another's model
• Simulated model risk committee sessions
• A ten-day project that builds a full governance pack for one model

Participants who attend all ten days and complete the labs and the governance project receive a CPD-accredited Certificate of Completion from Vision Reach Global Consultancy.

Alongside the certificate, each participant keeps the governance pack built during the course, including the model card, validation report and monitoring plan, as practical evidence of the controls they can apply to models in their own organisation.