AI, data and statistics

Machine Learning interview questions

Supervised learning, generalization and evaluation of predictive systems.

9 questions, each with a worked answer

What is overfitting? Fundamentals Concept
How would you split data when samples come from the same donor? Intermediate Scenario
Why might cross-validation performance not predict performance at a new hospital? Intermediate Scenario
How do feature leakage and label leakage differ? Intermediate Scenario
When is a simpler baseline model preferable to a complex model? Advanced Scenario
How would you test whether a model relies on batch artefacts? Advanced Scenario
How would you prevent leakage when a dataset includes multiple scans from each patient? Intermediate Troubleshooting
Model performance falls sharply after deployment to another hospital. How would you investigate? Intermediate Troubleshooting
Design a validation scheme for an ML model intended to predict outcomes in future calendar periods. Advanced Practical

About Machine Learning

Machine learning builds models that learn patterns from data to make predictions, such as a molecule's activity or a sample's class. Supervised methods learn from labelled examples, while unsupervised methods look for structure without labels. The central problem is generalisation: performing well on new data rather than memorising the training set.

In biology, small datasets, batch effects and leakage make careful validation especially important.

Fundamentals: 1
Intermediate: 5
Advanced: 3

What these questions test

Train ML models reproducibly: Version datasets and training procedures while preventing leakage.

References

Common pitfalls and recommended practices. scikit-learn User Guide. scikit-learn.org/stable/common_pitfalls.html.
scikit-learn.org/stable/common_pitfalls.
scikit-learn.org/stable/modules/generated/sklearn.model_selection.GroupKFold.html.
scikit-learn.org/stable/modules/calibration.html.

Practice material written by TheBench.Bio and checked against the references above, reviewed October 23, 2026. Written to practise with, not collected from real interviews.