Data Scientist Interview Questions

1.1What data science actually delivers - and the gap between expectation and reality in most organisations

1.2The data science workflow - problem framing, data, modelling, evaluation, and deployment

1.3Types of data science problems - classification, regression, clustering, NLP, and recommendation

1.4Working with stakeholders - translating vague business problems into tractable modelling challenges

2.1Probability theory - distributions, Bayes theorem, conditional probability, and their role in modelling

2.2Inferential statistics - confidence intervals, hypothesis testing, and when p-values mislead you

2.3Bayesian vs frequentist thinking - the practical difference and when each framework is more useful

2.4Experimental design - A/B testing, randomisation, power analysis, and avoiding confounding

3.1Supervised learning - regression and classification algorithms, their assumptions, and when to use each

3.2Unsupervised learning - clustering, dimensionality reduction, and finding structure in unlabelled data

3.3Model evaluation - train/test split, cross-validation, bias-variance trade-off, and choosing the right metric

3.4Overfitting, regularisation, and building models that generalise to new data

4.1Ensemble methods - random forests, gradient boosting, and why they work so well

4.2Neural networks - architecture, backpropagation, activation functions, and training deep models

4.3Natural language processing - text preprocessing, embeddings, transformers, and large language models

4.4Anomaly detection - statistical, distance-based, and model-based approaches and their trade-offs

5.1Feature selection - identifying the variables that matter and removing the ones that do not

5.2Feature engineering - creating new variables that improve model performance

5.3Handling missing data - imputation strategies and the risks of each approach

5.4Dealing with class imbalance - oversampling, undersampling, and cost-sensitive learning

6.1From notebook to production - the gap between a working model and a deployed, maintained system

6.2MLOps fundamentals - model versioning, monitoring, retraining, and the pipeline that keeps models working

6.3API design for ML - serving predictions at scale and managing latency

6.4Model monitoring - data drift, concept drift, and detecting when a model has stopped working

7.1Data pipeline fundamentals - how data flows from source to model and what can go wrong

7.2SQL for data scientists - the advanced querying skills needed to work independently with large datasets

7.3Working with big data platforms - distributed processing and working with data that does not fit in memory

7.4Data quality - validating inputs, handling schema changes, and building robust data pipelines

8.1Explaining models to non-technical audiences - interpretability, explainability, and honest communication

8.2Model limitations - how to communicate uncertainty, caveats, and what the model cannot do

8.3Presenting findings to executives - the key insight, the recommendation, and what to leave out

8.4Ethical responsibility - bias in models, fairness metrics, and the data scientist duty to flag problems

9.1Causal inference - moving beyond correlation to understanding what actually causes what

9.2Observational studies - propensity score matching, difference-in-differences, and managing confounding

9.3Multi-armed bandits - when to use exploration-exploitation approaches instead of fixed A/B tests

9.4Measuring business impact - how to attribute business outcomes to a model or intervention rigorously

10.1Technical interviews - statistics, ML algorithms, coding, and take-home projects

10.2Case study interviews - framing a problem, choosing an approach, and defending your modelling decisions

10.3Behavioural questions - failed models, stakeholder pushback, and impact you delivered

10.4Salary negotiation for data science roles, evaluating a data science team, and questions that reveal the real state of ML in the organisation