Top 35 data scientist interview questions

Hiring a great data scientist is not just about finding someone who can build models. The role requires a combination of statistical knowledge, coding ability, business understanding, and the judgment to turn messy data into decisions that drive results.

That makes interviewing data scientists particularly challenging. However, the right interview questions help uncover those capabilities. They reveal how candidates approach real-world problems, validate technical expertise, and distinguish practical experience from textbook knowledge.

In the World Economic Forum’s Future of Jobs Report 2025, big data specialists rank as the single fastest-growing job through 2030. When everyone is chasing the same scarce talent, your interview is the filter that decides who you actually land.

This guide gives you 35 questions across statistics, machine learning, hands-on coding, and real experience, plus a simple way to score the answers even if you are not a data scientist yourself.

TL;DR

• Test before you interview. A short skills assessment ranks your shortlist so the interview confirms a strong candidate instead of discovering a weak one.
• Use the 35 questions below as a menu, not a script. Pick 8 to 12 across statistics, machine learning, coding, and experience.
• In 2026, weight business judgment and modern tools (large language models, MLOps) alongside the classic statistics and machine-learning basics.
• Score answers with a simple 1 to 4 rubric so two interviewers reach the same call. We share one below (the Testlify Data Scientist Skills Scorecard).
• Getting it wrong is expensive: a bad hire can cost as much as $240,000, and teams without a standard process are 5 times more likely to make one.
• Watch for the candidate who knows every model but cannot say which metric the business cares about. Ask what their work changed, in dollars or hours.

Summarise this post with:

Chatgpt Gemini Claude Grok Perplexity

Why screen data scientists with a skills test first?

A pre-interview skills assessment gives you an objective read on whether a candidate can write working code and reason about data before you spend an hour in a room with them. It ranks your shortlist on evidence, so the interview becomes a confirmation, not a gamble.

Testlify helps teams evaluate candidates through skills tests, coding challenges, and role-specific assessments before the first interview. The goal is simple: ensure candidates have already demonstrated the core skills needed for the role before they reach your hiring team.

The value of skill tests is significant, especially when a hiring mistake is costly. Gallup puts the cost of replacing an employee at one-half to two times their annual salary, and for a senior data scientist, that is a six-figure mistake you would rather not repeat. A skills assessment helps reduce that risk by ensuring only qualified candidates move forward in the process.

When should you ask data scientist interview questions?

After the skills test, not before. Invite applicants to a short assessment first, then bring the top scorers in for the interview. That order saves your team from screening dozens of resumes by hand and protects strong candidates from a slow process.

SHRM research found that teams without a standardized hiring process are five times more likely to make a bad hire, so a consistent two-step flow is not bureaucracy; it is insurance. Use the interview to probe judgment, problem-solving, and how someone collaborates with non-technical teams.

What should a data scientist interview test in 2026?

Five areas, and the mix has shifted. Statistics and machine-learning fundamentals still matter most, but a growing share of real data work now touches large language models and the systems around them.

Stack Overflow’s developer survey found 76% of developers are using or planning to use AI tools, up from 70% the year before, so a data scientist who ignores them is already behind. Map your questions to the table below, and weight the rows that match the job.

General data scientist interview questions

These 25 questions cover the fundamentals every data scientist should own: machine learning, statistics, data handling, and how to evaluate a model. You will not ask all of them. Pick the six or seven that match the role, and follow up on any answer that sounds memorized. Check out Testlify’s Data Scientist test to screen these areas before the call.

1. What is the difference between supervised and unsupervised learning?

What a strong answer covers: Supervised learning trains on labeled data to predict an output; unsupervised learning finds structure in unlabeled data. A strong candidate gives real examples, like churn prediction for the first and customer segmentation for the second.

2. Explain the bias-variance tradeoff?

What a strong answer covers: Bias is error from a model that is too simple; variance is error from one that is too complex and memorizes noise. Look for the practical point: you tune model complexity and add data to find the balance, not a perfect fix.

3. What is overfitting, and how do you prevent it?

What a strong answer covers: The model nails the training data and fails on new data. Good answers name cross-validation, regularization, more data, and early stopping, and explain when each one helps rather than listing all four by rote.

4. What is cross-validation, and why use it?

What a strong answer covers: It splits the data into folds to estimate how a model will perform on data it has not seen. Look for k-fold and a clear reason it beats trusting a single train/test split.

5. Describe feature selection and why it matters?

What a strong answer covers: Cutting the inputs down to the ones that carry signal. Look for methods like forward selection, backward elimination, and Lasso, plus the payoff: less overfitting and a model people can actually interpret.

6. What is regularization, and when would you use it?

What a strong answer covers: A penalty added to the loss to keep coefficients from blowing up, which fights overfitting. Look for L1 (Lasso) versus L2 (Ridge) and why Lasso can zero out weak features.

7. How do you handle an imbalanced dataset?

What a strong answer covers: Resampling (oversampling the minority class or SMOTE), class weights, and metrics smarter than accuracy. The tell of a strong answer is knowing that raw accuracy is misleading when one class is rare.

8. Explain the difference between bagging and boosting?

What a strong answer covers: Bagging trains models in parallel on random subsets and averages them to cut variance; boosting trains them in sequence, each fixing the last one’s errors, to cut bias. Look for random forests and gradient boosting as examples.

9. What are common evaluation metrics for a classification model?

What a strong answer covers: Accuracy, precision, recall, F1, and ROC-AUC. The real signal is knowing which to use when, like recall for disease screening and precision for fraud flags, not reciting the whole list.

10. Explain how a decision tree works, with its pros and cons?

What a strong answer covers: It splits data on feature values into branches and leaves. Pros: easy to read and handles non-linear patterns. Cons: it overfits and is unstable. Look for the honest downside, not just the sales pitch.

11. What is gradient descent, and how does it work?

What a strong answer covers: An optimization method that steps down the loss surface to shrink error. Look for the learning rate, convergence, and the batch, stochastic, and mini-batch variants and their trade-offs.

12. How do you decide which chart to use for a dataset?

What a strong answer covers: Match the chart to the data type and the point you are making: bar charts for categories, histograms for distributions, scatter plots for relationships. Look for a candidate who thinks about the audience first.

13. What are best practices for an effective visualization?

What a strong answer covers: Pick the right chart, keep it simple, label clearly, use honest scales, and choose colors that stay readable for colorblind viewers. Look for someone who has seen a misleading chart and can name why it lied.

14. How would you visualize the distribution of a continuous variable?

What a strong answer covers: Histograms, box plots, and kernel density plots. Look for an explanation of what each one reveals about center, spread, and outliers, not just naming them.

15. What is a confusion matrix, and how do you use it?

What a strong answer covers: A table of true and false positives and negatives that summarizes a classifier’s performance. Look for the ability to read precision, recall, and F1 straight off it.

16. Walk me through how you would clean a messy dataset?

What a strong answer covers: Handle missing values, drop duplicates, fix data types, deal with outliers, and scale features. Look for a methodical order and the point that clean data, not a fancy model, usually drives the result.

17. How do you handle missing data?

What a strong answer covers: Impute it (mean, median, or a model), drop it, or use an algorithm that tolerates gaps. The strong answer first asks why the data is missing, because that decides which method is safe.

18. Explain exploratory data analysis (EDA) and why it matters?

What a strong answer covers: Summarizing and plotting a dataset to understand it before modeling. Look for summary statistics, distributions, and correlations, and the habit of looking at data before trusting it.

19. How do you detect and handle outliers?

What a strong answer covers: Spot them with box plots, Z-scores, or the IQR method, then decide to keep, transform, or remove based on whether they are real or errors. Look for judgment, not a blanket delete rule.

20. What is the significance of a p-value in hypothesis testing?

What a strong answer covers: It measures the evidence against the null hypothesis; a low p-value (under 0.05) is strong evidence to reject it. The best answers add what a p-value is not, since plenty of analysts misread it.

21. How do you assess whether a dataset is normally distributed?

What a strong answer covers: Visual checks (Q-Q plots, histograms) and tests like Shapiro-Wilk. Look for why normality matters for certain methods and what they would do when the data is clearly not normal.

22. Explain the Central Limit Theorem and why it is useful?

What a strong answer covers: The distribution of sample means approaches a normal curve as the sample grows, whatever the population looks like. Look for why that lets you use normal-based inference on real data.

23. What is multicollinearity, and how does it hurt a regression model?

What a strong answer covers: When predictors are highly correlated, which inflates standard errors and makes coefficients unstable and hard to interpret. Look for how they detect it, like a VIF check, and what they do about it.

24. How do you evaluate a regression model?

What a strong answer covers: R-squared, adjusted R-squared, MAE, MSE, and RMSE. The key is reading them in context, not reciting the list. Ask what they would actually report to a non-technical manager.

25. What is the difference between a Type I and a Type II error?

What a strong answer covers: A Type I error rejects a true null (a false positive); a Type II error fails to reject a false null (a false negative). Look for which one is costlier in a given business case and how they would trade them off.

Code-based data scientist interview questions

Talking about analysis is not the same as doing it. These five hands-on tasks (questions 26 to 30) show whether a candidate can actually write the code. Give them a shared editor or a take-home, and watch how they handle the parts that are easy to skip, like an empty list or a divide-by-zero.

Pro tip: Grade partial answers generously. A data scientist who writes clean, readable code and names the edge case they did not have time to handle is often a better hire than one who pastes a perfect but unexplained solution. How they think out loud matters more than a flawless first draft.

26. Write a Python function to return the mean and median of a list of numbers.

What to look for: Correct math, clean sorting for the median, and an empty-list guard so it does not crash.

def mean_median(numbers):
    if not numbers:
        return None, None
    mean = sum(numbers) / len(numbers)
    s = sorted(numbers)
    n = len(s)
    if n % 2 == 0:
        median = (s[n//2 - 1] + s[n//2]) / 2
    else:
        median = s[n//2]
    return mean, median

27. Write a SQL query to return the top 3 highest salaries from an employees table.

What to look for: Correct use of ORDER BY and LIMIT, and a candidate who asks whether ties should count.

SELECT salary
FROM employees
ORDER BY salary DESC
LIMIT 3;

28. Write a Python function to remove duplicates from a list while keeping order.

What to look for: Comfort with sets and lists, and a loop that preserves the original order instead of just calling set().

def remove_duplicates(lst):
    seen = set()
    result = []
    for item in lst:
        if item not in seen:
            result.append(item)
            seen.add(item)
    return result

29. Use Pandas to filter the rows of a DataFrame where age is greater than 30.

What to look for: Everyday Pandas fluency and boolean indexing without overcomplicating it.

import pandas as pd

df = pd.DataFrame({
    'name': ['Alice', 'Bob', 'Charlie', 'David'],
    'age': [25, 35, 30, 40]
})

filtered_df = df[df['age'] > 30]

30. Write a function to count the frequency of each word in a string.

What to look for: String handling and a dictionary (or Counter) for the count, plus a thought about case and punctuation.

def word_frequency(text):
    frequency = {}
    for word in text.lower().split():
        frequency[word] = frequency.get(word, 0) + 1
    return frequency

Which questions reveal a candidate’s real experience?

The ones that make someone tell a story. Theory shows what a candidate knows; experience questions show what they have actually done when a project went sideways. Ask these five (questions 31 to 35) and listen for specifics, real datasets, real stakeholders, and a clear account of what their work changed.

31. Describe a time you explained a complex data finding to a non-technical stakeholder. What landed, and what did not?
32. Tell me about the most challenging data project you shipped. What was the hardest blocker, and how did you get past it?
33. How do you prioritize when three teams all want analysis from you in the same sprint? Give a real example.
34. Tell me about a time your analysis pointed the business the wrong way, or a model underperformed in production. What did you do next?
35. How do you keep up with a field that moves this fast? Name what you read, build, or follow.

Vague answers are the tell. A senior data scientist names the dataset, the metric, the constraint, and the trade-off they accepted. Someone padding their resume stays abstract because there is no real project underneath the words.

How do you score answers without a data science background?

Use a rubric so you are judging evidence, not vibes. You do not need to build models yourself to spot a strong answer; you need a consistent scale and a few things to listen for. The Testlify Data Scientist Skills Scorecard rates each candidate from 1 to 4 across the five areas in the table above, and it weights communication and business sense as heavily as the math, because that is where most data hires quietly fail on the job.

• 1 (weak): Recites a definition, no examples, freezes on follow-ups.
• 2 (developing): Knows the theory but cannot connect it to a real project or a business goal.
• 3 (Strong): Explains clearly, gives concrete examples, names the trade-offs.
• 4 (Exceptional): Teaches you something, including when an approach fails and what they would try instead.

Have two interviewers score independently, then compare. When their numbers diverge by more than a point, that gap is usually where the real signal hides, so talk it through before you decide.

Pair this with the skills-test score, and you have an evidence-based call, not a popularity contest. For more on why tested skills beat resumes, see our take on prioritizing skills over degrees and how skill testing leads to better hiring decisions.

Key takeaway: The best predictor of a data scientist’s on-the-job success is not the degree or the Kaggle rank. It is whether they can connect a model to a decision and explain why it might be wrong. Test for that first, then interview to confirm it. A bad hire here can run up to $240,000, so the teams that screen on real skills are the ones that stop paying for it.

Hire data scientists with confidence

Start every data hire with proof of skill. Send candidates a role-based data scientist assessment, mix in a live coding project for the shortlist, and walk into the interview already knowing who can do the work.

Start free with Testlify, or book a demo to see how to build the test-then-interview flow for your team.

Frequently asked questions (FAQs)

Reuben

Linkedin profile

Content Writer

Reuben John is a B2B content writer focused on HR and recruitment. His work explores hiring trends, skills-based recruitment, talent assessment, and the technologies shaping how companies find and hire talent.