Question 1
What decision would a model support in this job, and what would you refuse to automate?
Approach. Name the decision, the label you can actually collect, and the human who still owns the call.
Data scientist interview questions from a job description
Turn a data scientist job description into likely modeling, experiment, and stakeholder questions, grouped by the skill they test.
Matched in your browser from the requirements in the posting and Luna’s interview-question hub. These are practice prompts, not a claim that this wording will be asked.
Requirements in the posting
Data Scientist
Help product teams decide what to build with measurement and models.
You will:
- Analyze behavior in SQL and Python, including pandas where it is the right tool
- Design experiments and state the hypothesis, the metric, and the decision
- Build machine learning models when a decision needs a prediction, and say when it does not
- Explain uncertainty to stakeholders who want a yes or no
- Own the analysis from question through the metric the team will watch
We care that you can tell a product partner what the data cannot support.
Help product teams decide what to build with measurement and models.
You will:
- Analyze behavior in SQL and Python, including pandas where it is the right tool
- Design experiments and state the hypothesis, the metric, and the decision
- Build machine learning models when a decision needs a prediction, and say when it does not
- Explain uncertainty to stakeholders who want a yes or no
- Own the analysis from question through the metric the team will watch
We care that you can tell a product partner what the data cannot support.
Likely questions
Role
Machine learning
Build machine learning models when a decision needs a prediction, and say when it does not
Question 2
How would you know a model that looks good offline is hurting users?
Approach. Say the online check and the slice — a group of users or inputs — you would watch first.
Coding
Python
Analyze behavior in SQL and Python, including pandas where it is the right tool
Question 3
How would you structure a Python module for the core task in this posting so someone else can change it?
Approach. Name the function boundary and the test you would write before adding libraries.
Question 4
Tell me about a time a Python script or service was correct and still too slow.
Approach. Say what you measured, the change, and the cost you left in place.
Role
Experiments and statistics
Design experiments and state the hypothesis, the metric, and the decision
Question 5
What would you experiment on in the first month of this job, and what metric would decide it?
Approach. One change, one metric, and the confound you would not ignore.
Question 6
Tell me about a result that was statistically neat and still the wrong product call.
Approach. Say who the metric left out and the decision you made anyway.
Coding
SQL and data modeling
Analyze behavior in SQL and Python, including pandas where it is the right tool
Question 7
What tables would you need for the core object in this posting, and what is the grain of each?
Approach. Name the entity, the key, and one query you must keep fast.
Question 8
Write the shape of a query this team would run every day. Where would it get slow?
Approach. Say the join, the filter, and the index or rewrite you would try first.
Behavioral
Collaboration
Explain uncertainty to stakeholders who want a yes or no
Question 9
Tell me about a disagreement with someone who did not report to you. What did you do with their objection?
Approach. Name the objection, what you changed, and what you did not. Do not flatten it into “we aligned.”
Question 10
How do you bring a partner along when this posting’s work depends on their team?
Approach. Say the ask, the artifact you left them, and what you refused to take over.
What to prepare
- Prepare one model story: the decision it informed, the metric, and where it was wrong.
- Solve one small task in Python while saying the cost of the obvious version.
- Write one experiment: the metric, the unit of randomization, and the result that would stop you.
- Practice one join out loud: the grain of the result, then the filters.
- Have one story where someone disagreed and you kept the working relationship.
- Mark the three requirements you cannot yet support with a story.
- Say each answer out loud once, in under two minutes.
Data scientist postings mix three interviews: a technical screen in Python or SQL, a modeling or experiment discussion, and a stakeholder story. The bullets tell you the mix. “A/B” and “hypothesis” mean they will ask what would change your mind. “Model” means they will ask where the model was wrong online. “Partner with product” means the result has to be a decision, not a notebook.
Read the metric carefully. If the posting names a business number, your stories need that kind of number. If it names model quality only, they will still ask who feels a false positive.
Prepare one analysis that moved a decision and one model or test you would not ship. The sample posting is a product data science role.
Sample job description
The posting below is already in the tool. The questions are the ones its requirements point to. Replace it with a posting you are actually interviewing for.
Related: Netflix role-specific questions · Google interview questions · SQL interview pacing