Home/Interview prep/Data Science Fundamentals

Walk into the interview already prepared.

Timed mock tests and model answers written by working engineers, for every level from fresher to lead. Pick any of our 75 courses below.

Topic: AI & Machine Learning. Questions are shared by related courses. View the Data Science Fundamentals course →

All Data Science Fundamentals interview questions at a glance+

Fresher (0–3 years): Junior Data Scientist / ML Engineer

  1. Explain the bias–variance trade-off.
  2. How do you handle missing values?
  3. Walk me through a machine learning project end to end.
  4. What is the difference between AI, machine learning and deep learning?
  5. Predicting house prices from features is which type of problem?
  6. Grouping customers without labels is…
  7. A model scores 99% on training data but 60% on new data. This is…
  8. For a fraud model where fraud is 1% of cases, accuracy is misleading. Better metric?
  9. Why split data into training and test sets?
  10. What is one-hot encoding?
  11. What does a large language model (LLM) mainly do?
  12. What does cross-validation provide?

Mid-level (3–8 years): Data Scientist / ML Engineer

  1. Your model performs well offline but poorly in production. Why might that be?
  2. How do you choose evaluation metrics with the business?
  3. Explain how you would fine-tune a pre-trained model.
  4. How do you do feature engineering for a churn model?
  5. A feature computed with information from the future leaks into training. This is…
  6. Gradient boosting (e.g. XGBoost) builds…
  7. Which helps a model generalise by penalising large weights?
  8. Class imbalance can be handled by…
  9. Transfer learning means…
  10. ROC-AUC of 0.5 means the model is…
  11. What is the attention mechanism in transformers?
  12. Why scale features for k-NN or SVM?

Experienced (8–15 years): Senior / Lead ML Engineer

  1. How would you build and lead an ML team?
  2. When would you use prompting, RAG or fine-tuning for an LLM use case?
  3. How do you handle bias and fairness in a model?
  4. An ML system costs too much to run. How do you cut cost?
  5. Best way to prove a new model improves the business?
  6. A feature store mainly prevents…
  7. For a domain chatbot over private documents, the usual first approach is…
  8. Responsible AI work includes…
  9. Which is a key risk when deploying ML in regulated decisions (e.g. lending)?
  10. Quantisation of a model mainly…
  11. When should you choose a simpler model over a complex one?
  12. Most ML project failures come from…