Technical screening questions
Ask for one example per question and listen for specifics.
- How do you choose between models? A good answer: Baseline first, then validation and trade-offs.
- How do you design an A/B test? A good answer: Hypothesis, sample size and metrics.
- How do you handle class imbalance? A good answer: Metrics, resampling and thresholds.
- How do you explain a model to a non-technical stakeholder? A good answer: Plain language, impact and uncertainty.
- Describe a project that went to production. A good answer: Ownership beyond the notebook.
Quick screening questions
Use these when you have only a few minutes.
- Describe a model that changed a business decision.
- How do you design an A/B test?
- How do you explain a model’s limits to a non-technical stakeholder?
What separates the levels
Match the level to the requirement before you submit.
- Junior: analyzes and builds models with guidance.
- Mid: owns projects.
- Senior: sets analytics strategy and mentors.
Red flags
Any of these is a reason to probe further.
- Cannot explain the metric used.
- Only course projects.
- No communication examples.
Tools you should hear mentioned
A data scientist with real experience usually names several of these without prompting.
- Python or R
- SQL
- scikit-learn
- Jupyter
- Statistics
- Visualization libraries
Run the call well
Ask for the last project first, then dig into the second or third question. Take notes on versions, team size and outcomes, and record them with the submission so the vendor’s later technical interview does not surprise you.
Frequently asked questions
How long should a data scientist screening call take?
Fifteen to twenty minutes is enough to cover the project, three technical questions and availability, rate and work authorization.
Do recruiters need to know the technology?
Working vocabulary is enough. Use the sourcing guide for the role and listen for specifics instead of judging technical depth yourself.