Visa Data Scientist interviews reported by candidates combine SQL, Python, and machine learning questions with resume walkthroughs and behavioral discussion. Here is what the stages looked like and how to prepare.
$155K
Avg. Base Comp
$180K
Avg. Total Comp
3-6 rounds
Typical Rounds
4 weeks
Process Length
Candidates describe the Visa Data Scientist interview as a mix of hands-on coding and conversation about your own work. SQL is the most consistent technical thread: one candidate faced an online assessment with a SQL and a Python question, another a screen with a three-part SQL problem (GROUP BY, average transaction amount, LEAD/LAG), and another a round on window functions, aggregates, and dynamic variables. Python was reported at an easy-to-medium level, including a Fibonacci problem and Pandas data manipulation.
Machine learning questions tended to be applied. One screen included a case on identifying high-intent users, with XGBoost, overfitting, and cross-validation discussed. Another candidate was asked about classification metrics such as precision, recall, and ROC curves, plus a statistics question on doubling every data point in a linear regression. A separate loop reportedly covered end-to-end modeling from feature engineering to monitoring and retraining, and a RAG pipeline design question.
Expect to talk about your past projects. Candidates mention resume walkthroughs with hiring managers and questions about the business requirement behind a project. One candidate also got a consulting-style question about using Visa data to inform opening a retail business. Candidate accounts differ on structure, from a short recruiter, SQL, and hiring manager sequence to a five-round loop plus recruiter call, so confirm the format with your recruiter. Outcomes in the reports were mostly no offer or still waiting, and one candidate described scheduling with little notice.
Synthesized from 6 candidate reports by our editorial team.
Had an interview recently?
Share your experience. Unlock the full guide.
Real interview reports from people who went through the Visa process.
There were 5 rounds + 1 recruiter round. 1st Round with HM(1 hr) - Resume walkthrough and ML case study Onsite loop
Round 1 - Again Resume walkthrough and ML System Design Round 2 - Resume walkthrough and Model Deployment(full lifecycle) Round 3 - Behavioral Round 4 - Resume walkthrough and Python coding question
Questions asked: It was mostly related to how would you create a model end to end - starting with feature engineering to model monitoring and retraining. Was asked to design a RAG pipeline Python medium level question in coding
Share your own interview experience to unlock all reports, or subscribe for full access.
Sourced from candidate reports and verified by our team.
Topics based on recent interview experiences.
Featured question at Visa
Write a query to get the total three-day rolling average for deposits by day
| Question | |
|---|---|
| Closest SAT Scores | |
| Employee Salaries | |
| Top Three Salaries | |
| Alphabet Sum | |
| Encoding Categorical Features | |
| Total Spent on Products | |
| Bagging vs Boosting | |
| Sum to N | |
| Size of Joins | |
| Fewer Orders | |
| Resumable Fact Table Load | |
| Employees Before Managers | |
| Hurdles In Data Projects | |
| Production Model Monitoring | |
| RAG Strict Source Control | |
| Count Transactions | |
| Concurrent LLM Serving | |
| Modifying a Billion Rows | |
| Slow SQL Query | |
| Implementing the Fibonacci Sequence in Three Different Methods | |
| Filling Supermarket Bag | |
| Swap Variables | |
| User Event Data Pipeline | |
| Delivery Assignments | |
| Mouse Search | |
| Location Feature Sharing | |
| Overfit Avoidance | |
| Solo ML Deployment | |
| Three Indexes Adding Zero |
Synthesized from candidate reports. Individual experiences may vary.
Candidates report different starts. One describes an online assessment with one SQL and one Python question, followed by a recruiter call. Others report a recruiter conversation about their background first. Timing and format may vary by team, so ask early what to expect.
Candidates report SQL at easy-to-intermediate difficulty, including window functions, aggregates, and LEAD/LAG, alongside easy Pandas or Python tasks. One screen added an ML case on identifying high-intent users, with overfitting and cross-validation questions. Short notice for this round was reported in one account.
Candidates report hiring manager discussions centered on past projects and analysis approach, with one spending roughly the first twenty minutes on this. Questions about career goals and a short Python test at the end were reported by one candidate. Prepare specific projects with measurable outcomes.
Accounts of the later stage vary. One candidate describes a loop of four rounds covering ML system design, model deployment across the full lifecycle, behavioral questions, and Python coding, with resume walkthroughs in several. Another candidate described a super day with business acumen, ML concepts, and experimental design interviews.
Reported outcomes were no offer for three candidates and a pending decision for two. One candidate received an automated rejection about two weeks after the final conversation with no feedback. Another reported a total of about four weeks from application to final round.
