Credit Karma Data Scientist interview typically runs 5 rounds: recruiter screen, hiring manager screen, coding/ML round, virtual onsite, final functional interview. It usually takes about 2-3 months and is notably recommendation-systems heavy.
$132K
Avg. Base Comp
$237K
Avg. Total Comp
5
Typical Rounds
2-3 months
Process Length
We’ve seen Credit Karma evaluate data scientists through a very specific lens: can you connect modeling choices to product outcomes in a financial marketplace? Across candidate reports, the deepest conversations kept circling back to recommendation systems, from two-tower architectures and embedding tradeoffs to when kNN makes sense versus ANN. That tells us the team is not just checking whether you know the vocabulary; they want to hear how you reason about retrieval, ranking, and business metrics as one system, especially in a consumer finance context where recommendations have to be both useful and defensible.
A recurring theme is that the process looks standard on paper but often becomes more exploratory in the room. Multiple candidates reported that rounds labeled as behavioral or general ML shifted into live case discussions, including binary classification scenarios and recommendation-system design tied to marketing campaigns. That means the real signal is often how quickly you can adapt your thinking when the interviewer pushes beyond the expected script. We’ve also seen that the company seems comfortable with a conceptual depth over pure coding grind: LeetCode-style questions show up, but they’re rarely the whole story.
The other pattern we’d flag is that interviewers appear patient and engaged, yet the process itself can feel loosely coordinated. Candidates repeatedly mentioned unclear expectations between rounds and a lot of follow-up to get basic logistics. In practice, that means the strongest candidates are the ones who can stay crisp and structured even when the conversation is a bit messy. At Credit Karma, the bar seems to be less about perfect polish and more about whether you can make sound tradeoffs, explain them clearly, and keep the discussion anchored to user impact.
Synthetized from 2 candidates 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 Credit Karma process.
The position was for a senior DS role, and the process ended up being a lot longer than I expected. It started with a recruiter screen, then a hiring manager conversation, then a leet coding and basic ML round, and finally a virtual onsite with two technical interviews. The first onsite round focused on recommendation systems, and the second was broader ML plus recommendation systems and system design. The interviewers themselves were respectful and patient, and they really did listen to my answers instead of rushing me, which I appreciated.
What made the process frustrating was the lack of clarity between rounds. I often didn’t know what the next step would be, what each interview was supposed to cover, or even when it would happen, so I had to follow up multiple times just to get basic scheduling and prep information. There were also several reschedules because responses were delayed, and the email threads got messy with a lot of HR people involved. The onsite was virtual, and one interviewer was on-call and had to jump into Slack messages a few times during the interview, which they warned me about at the start, but it still made the round feel a bit distracting.
On the technical side, I was asked a LeetCode-style dynamic programming question, some ML basics, and then a lot of depth around recommendation systems. The recsys discussion was flexible and depended on how I answered, but it went into things like two-tower models, different embedding choices and when to use them, how to think about the loss function when training, and how all of that connects back to business metrics. I also got a kNN vs ANN question and had to explain why you’d use each one. Overall it felt more like a deep conceptual interview than a pure coding grind, but it still required solid fundamentals and the ability to reason through tradeoffs on the spot. After about three months and five rounds, I got an automated rejection email.
Prep tip from this candidate
Be ready to go deep on recommendation systems rather than just naming models: practice explaining two-tower architectures, embedding tradeoffs, loss design, and how those choices map to business metrics. Also review a dynamic programming LeetCode problem and be able to clearly compare kNN vs ANN and when each makes sense.
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 Credit Karma
How would you set up this test?
| Question | |
|---|---|
| The Brackets Problem | |
| Type-ahead Search | |
| Testing Price Increase | |
| Offer Matching API Design | |
| A/B Testing a Checkout Button Change | |
| Check Matching Parentheses | |
| Netflix Price | |
| Empty Neighborhoods | |
| 2nd Highest Salary | |
| Rolling Bank Transactions | |
| Comments Histogram | |
| Closest SAT Scores | |
| Top Three Salaries | |
| Merge Sorted Lists | |
| Cumulative Distribution | |
| Experiment Validity | |
| String Shift | |
| Last Transaction | |
| Bagging vs Boosting | |
| Alphabet Sum | |
| Prime to N | |
| Paired Products | |
| Bank Fraud Model | |
| P-value to a Layman | |
| Swipe Precision | |
| Unique Work Days | |
| Over-Budget Projects | |
| Third Purchase | |
| Top 3 Users |
Synthesized from candidate reports. Individual experiences may vary.
An initial conversation with recruiting to review your background, role fit, and basic logistics. In the experiences, this was the first step before any technical interviews, though communication and scheduling could be slow.
A conversation with the hiring manager focused on your experience and technical fundamentals. Candidates reported ML basics, background questions, and an early check on fit for the senior data scientist role.
A technical interview covering LeetCode-style coding and core machine learning concepts. One candidate saw a dynamic programming problem, while another described medium-level coding alongside basic ML questions.
A virtual onsite with multiple technical interviews. Reported topics included ML Ops, broader ML fundamentals, case studies, and recommendation systems, with some rounds going deep on two-tower models, embeddings, loss functions, and business metrics.
A final interview that was expected to be behavioral but could shift into a technical case discussion. Candidates reported recommendation-system case work, ML system design for marketing or binary classification use cases, and discussion with a director or hiring manager.