Data

• SQL for analytical workloads: joins, window functions, aggregation at scale
• Data quality assessment and honest handling of missingness
• Feature engineering and the leakage traps that inflate offline metrics

Modelling

• Regression and classification workflows with proper validation splits
• Regularisation, hyperparameter search and overfitting diagnosis
• Clustering and dimensionality reduction, and their instability
• Interpretability with SHAP and partial dependence for stakeholder trust

Inference

• A/B test design, power analysis and stopping rules
• Quasi-experimental methods when randomisation is impossible
• Distinguishing correlation from a claim that will drive a decision

Delivery

• Visualisation that supports a decision rather than decorating a deck
• Communicating uncertainty to non-technical stakeholders
• Packaging analysis for reproducibility and handover to engineering