AI & Data Science

Translate a business question into an analytical task and begin working with data in Python.

• The data science lifecycle: problem, data, analysis, modelling and communication.
• Python variables, collections, functions, loops and reusable scripts.
• Notebooks, files, libraries and reproducible working environments.
• Choosing a target, defining success and recognising unsuitable use cases.

Practical exerciseExplore a sample dataset and write a project brief with a measurable analytical question.

Use summary statistics and visual exploration to understand patterns before building a model.

• Distributions, averages, variation, outliers and missing values.
• Probability, sampling and introductory hypothesis-testing concepts.
• NumPy and pandas for selection, filtering, grouping and aggregation.
• Charts, relationships and the difference between correlation and causation.

Practical exerciseProduce an exploratory notebook with data-quality findings, visualisations and clearly stated observations.

Prepare reliable model inputs while avoiding information leakage.

• Cleaning duplicates, inconsistent types and invalid records.
• Imputation, categorical encoding, scaling and transformations.
• Train, validation and test splits, including time-aware splits where needed.
• Feature selection, preprocessing pipelines and reproducibility.

Practical exerciseBuild a repeatable preparation pipeline and explain which transformations are learned only from training data.

Learn how models make predictions and compare them against a useful baseline.

• Regression and classification problems; linear and logistic models.
• Decision trees, random forests and boosting concepts.
• Training, hyperparameters, cross-validation and overfitting.
• MAE, RMSE, precision, recall, F1 and choosing metrics for the task.

Practical exerciseTrain and compare baseline and tree-based models, documenting the trade-offs using held-out data.

Explore structure in unlabelled data and explain model behaviour carefully.

• Clustering, distance measures and choosing meaningful features.
• Dimensionality reduction and principal component analysis concepts.
• Anomaly-detection use cases and limitations.
• Feature importance, error analysis and communicating uncertainty.

Practical exerciseExplore customer segments or unusual records and explain why the results require domain validation.

Understand where neural models fit and how text becomes a usable model input.

• Neurons, layers, activations, loss and optimisation concepts.
• Training and validation curves; regularisation and data requirements.
• Text preprocessing, tokenisation, embeddings and text classification.
• Pretrained models, transfer learning and the relationship to generative AI.

Practical exerciseRun a guided text-classification experiment and compare results with a simpler baseline.

Plan how a model can be used, monitored and reviewed after an experiment.

• Saving models, inference inputs and simple API delivery concepts.
• Experiment tracking, versions and reproducible results.
• Data drift, monitoring, retraining triggers and human oversight.
• Privacy, bias, explainability and documenting model limitations.

Practical exercisePrepare a model card and an inference demonstration with input validation and a monitoring plan.

Complete an end-to-end analysis and explain the evidence behind your decisions.

• Problem statement, data dictionary and exploratory findings.
• Preparation pipeline, baseline, model comparison and evaluation.
• Results, limitations and a practical recommendation.
• Python, statistics, machine learning and project-based interview questions.

Practical exercisePresent a reproducible project notebook, evaluation report and model demonstration using a suitable sample dataset.