Skip to content

Data Scientist: what the role asks for

Builds statistical and machine-learning models to answer questions or power product features. Designs experiments, prepares features and evaluates models for accuracy and fairness. Communicates uncertainty and limits of results to non-specialists.

Skill profile: Karnamo curated v1 (original, not ESCO-derived). It describes a typical role; individual employers differ, so always check the job description in front of you.

Core skills

Expected from day one. Gaps here usually decide whether you get to the interview.

  • Python

    target: deep / can teach others

    General-purpose language common in scripting, back-end services, automation and data work.

  • Statistics

    target: deep / can teach others

    Probability, inference and experimental reasoning for sound conclusions from data.

    • Work through confidence intervals, p-values and power on a simulated dataset and explain each result in plain language (practice, free)
  • Machine Learning

    target: deep / can teach others

    Building, evaluating and deploying predictive models.

    • scikit-learn user guide (reading, free)
    • Train and evaluate a baseline model on a public dataset, documenting leakage risks and error analysis (project, free)
  • SQL

    target: solid, independent

    Query language for relational databases: filtering, joins, aggregation and schema design.

  • Data Analysis

    target: deep / can teach others

    Exploring data, finding patterns and turning them into decisions.

    • Analyse a public dataset end to end and write a one-page summary with a clear recommendation (project, free)

Important skills

Expected within the first year; strong candidates show some already.

  • A/B Testing & Experimentation

    target: deep / can teach others

    Designing and interpreting controlled experiments for product decisions.

    • Design an experiment on paper: hypothesis, primary metric, sample size, guardrails and decision rule (practice, free)
  • Data Visualization

    target: solid, independent

    Presenting data clearly in charts and dashboards.

    • Rebuild a misleading chart honestly and write down which choices you changed and why (practice, free)
  • Git

    target: solid, independent

    Distributed version control, branching strategies and collaboration workflows.

  • Presentation Skills

    target: solid, independent

    Structuring and delivering talks and demos for different audiences.

    • Record a ten-minute presentation, watch it back and note three concrete improvements (practice, free)
  • Stakeholder Management

    target: solid, independent

    Aligning expectations and communication with people who influence or are affected by your work.

    • Map the stakeholders of a current project by influence and interest and plan how to communicate with each (practice, free)

Nice to have

Differentiators that help but rarely block.

  • Data Engineering

    target: solid, independent

    Building pipelines and warehouses that move and prepare data reliably.

    • Build a small pipeline that ingests a public API daily, validates it and loads it into a warehouse table (project, free)
  • Data Protection & GDPR

    target: solid, independent

    Handling personal data lawfully, including privacy by design and data-subject rights.

Sample interview questions

  1. Technical / domain

    Explain overfitting and how you would detect and reduce it.

    Accurately describes train versus validation behaviour, with options like regularisation, more data and simpler models.

  2. Technical / domain

    Explain the difference between correlation and causation and how you would test for a causal effect.

    Shows an accurate grasp of confounding and randomised experiments, plus alternatives when experiments are not possible.

  3. Coding or case study

    Here is a dataset with a surprising result. Explore it and tell me what you would check before trusting it.

    Checks for data quality, selection effects and confounding before offering explanations, and states confidence honestly.

  4. Technical / domain

    How do you choose an evaluation metric for a classification model with imbalanced classes?

    Explains why accuracy misleads, then ties precision, recall or other metrics to the business cost of each error type.

  5. Final

    How do you make sure the insights you produce are actually used?

    Shows early involvement of decision makers, clear recommendations, follow-up and feedback loops.

  6. Panel

    Present an analysis you did and explain its limitations to a mixed technical and non-technical audience.

    Shows you tailor depth to the audience and state limitations and confidence clearly.

All Data & analytics interview questions · Pick your own round and seniority

Related career moves

Roles whose skill profiles overlap most with this one. Overlap is a starting point for a conversation, not a recommendation.

Compare your own CV against Data Scientist job postings with a transparent rubric.

Try the job-fit analyzer