Career1. Explain overfitting and how you prevent it.
Why they ask: Core ML understanding.
How to answer:
- The model learns noise in training data and fails on new data.
- Prevent it with cross-validation, regularisation, simpler models and more data.
Data Scientist Interview Questions (ML, Stats, SQL) with Tips
Data science interviews test statistics, modelling judgment and, above all, whether your work changed a decision.
- Explain overfitting and how you prevent it.
- How do you choose an evaluation metric?
- Walk me through a model you deployed.
- How would you design an experiment to test a new feature?
- Explain a p-value to a non-technical manager.
- How do you handle messy or biased data?
2. How do you choose an evaluation metric?
Why they ask: Matching models to the business.
How to answer:
- Start from the cost of mistakes: false positives vs false negatives.
- Accuracy is misleading on imbalanced data; use precision, recall, AUC or business metrics.
3. Walk me through a model you deployed.
Why they ask: End-to-end experience.
How to answer:
- The business problem and data.
- Features, model choice and validation.
- Deployment, monitoring and impact.
4. How would you design an experiment to test a new feature?
Why they ask: Causal thinking.
How to answer:
- Hypothesis, primary metric and guardrails.
- Sample size and randomisation.
- How you would read the result.
5. Explain a p-value to a non-technical manager.
Why they ask: Communication.
How to answer:
- How surprising the result would be if there were really no effect.
- Pair it with the size of the effect and what it means for the business.
6. How do you handle messy or biased data?
Why they ask: Data judgment.
How to answer:
- Understand how the data was collected.
- Clean, document assumptions, and check for bias against groups.
More interview guides
Any job title works in CareerCue — these are just a starting point. See all fields →
Mind going blank in the real interview?
CareerCue works for any job in any field. It listens to the question and shows a short cue based on your resume and the role, so you can answer in your own words. Free to try, no credit card.
Try CareerCue free →