Zum Inhalt springen

Data Scientist: Was die Rolle erfordert

Einige Inhalte auf dieser Seite (Kompetenzen, Fragen) sind derzeit nur auf Englisch verfügbar.

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.

Kompetenzprofil: Karnamo curated v1 (original, not ESCO-derived). Es beschreibt eine typische Rolle; einzelne Arbeitgeber unterscheiden sich. Prüfe daher immer die konkrete Stellenbeschreibung.

Kernkompetenzen

Ab dem ersten Tag erwartet. Lücken entscheiden hier oft darüber, ob du zum Interview eingeladen wirst.

  • Python

    Zielniveau: vertieft / kann andere anleiten

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

  • Statistics

    Zielniveau: vertieft / kann andere anleiten

    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, kostenlos)
  • Machine Learning

    Zielniveau: vertieft / kann andere anleiten

    Building, evaluating and deploying predictive models.

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

    Zielniveau: solide, selbstständig

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

  • Data Analysis

    Zielniveau: vertieft / kann andere anleiten

    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, kostenlos)

Wichtige Kompetenzen

Im ersten Jahr erwartet; starke Bewerber bringen einige bereits mit.

  • A/B Testing & Experimentation

    Zielniveau: vertieft / kann andere anleiten

    Designing and interpreting controlled experiments for product decisions.

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

    Zielniveau: solide, selbstständig

    Presenting data clearly in charts and dashboards.

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

    Zielniveau: solide, selbstständig

    Distributed version control, branching strategies and collaboration workflows.

  • Presentation Skills

    Zielniveau: solide, selbstständig

    Structuring and delivering talks and demos for different audiences.

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

    Zielniveau: solide, selbstständig

    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, kostenlos)

Wünschenswert

Zusätzliche Stärken, die helfen, aber selten eine Voraussetzung sind.

  • Data Engineering

    Zielniveau: solide, selbstständig

    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, kostenlos)
  • Data Protection & GDPR

    Zielniveau: solide, selbstständig

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

Beispielhafte Interviewfragen

  1. Technik / Fachgebiet

    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. Technik / Fachgebiet

    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. Programmierung oder Fallstudie

    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. Technik / Fachgebiet

    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. Abschlussgespräch

    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. Gesprächsrunde mit mehreren Beteiligten

    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.

Alle Interviewfragen für Data & analytics · Gesprächsrunde und Erfahrungsstufe selbst wählen

Verwandte berufliche Wechsel

Rollen, deren Kompetenzprofile sich mit diesem am stärksten überschneiden. Die Überschneidung ist ein Gesprächseinstieg, keine Empfehlung.

Vergleiche deinen Lebenslauf anhand eines transparenten Rasters mit Stellenanzeigen für Data Scientist.

Passungsanalyse ausprobieren