Capital One Interview Questions — Page 27

Capital One·Machine Learning Engineer·Online Assessment (OA)

Jun 2026

Capital One ML Engineer OA, four independent coding problems with no behavioral component. The problems ranged from pretty approachable to genuinely annoying to implement correctly, especially that second one with the repeated subtraction logic.

Given an array of daily website visit counts and a target number, return the earliest day index where the cumulative visits reach or exceed that target.
Simulate a repeated subtraction process on an array: repeatedly find the first positive value, subtract it from all following elements until you hit something smaller, accumulate it into a result, and repeat until no positives remain.
Place a series of labeled blocks with fixed shapes onto a grid, choosing the topmost then leftmost valid position for each, and return the final grid state.
Given an array of house heights where you can only increase values, find the minimum number of increments to make the heights form either a strictly increasing or strictly decreasing sequence with step size exactly 1.

“Straightforward prefix sum scan.”

View Post

Capital One·Machine Learning Engineer·Onsite - Behavioral / Leadership

Jun 2026

Three behavioral questions for an ML Engineer role at Capital One. Nothing technical, just back-to-back stories about handling hard situations. Felt a bit repetitive by the third one but I got through it.

Tell me about a time you faced a significant challenge. What made it hard, and how did you push through?
Describe a situation where you were blocked on something but still managed to deliver on time. How did you get yourself unblocked?
Tell me about something you accomplished that most of your peers struggled with. What gave you the edge?

“This is the warmup question that somehow still trips people up.”

View Post

Capital One·Software Engineer·Technical Phone Screen

Jun 2026

Capital One data engineer interview with a role-play scenario where you have to explain model degradation to a non-technical PM. Pretty involved for a data engineering role, felt more like a data science communication exercise than anything I expected.

You are a data scientist. A PM tells you the model performed well on validation but degraded after going live. How do you explain what happened, walk through how you'd debug it, and propose fixes to prevent it from happening again? Assume the PM is non-technical and cares about user impact and timelines.

“The role-play format threw me a bit.”

View Post

Capital One·Software Engineer·Technical Phone Screen

Jun 2026

Capital One Applied Researcher interview with a single coding-style question. Pretty straightforward problem, nothing that felt like it was testing deep research instincts.

You're given an array of daily website visit counts. Find the index of the first day where the running total of visits meets or exceeds a given target value.

“Basically just iterate and keep a running sum, return the index when you cross the threshold.”

View Post

Capital One·Data Scientist·Technical Phone Screen

Jun 2026

Capital One data scientist interview that went deep on ML algorithm selection for a tabular prediction problem. The whole thing felt like a single extended case study with a lot of follow-ups layered on top of each other.

You need to pick an algorithm for predicting arrival delay from a 500k-row dataset with 120 mixed features, non-linear interactions, sub-100ms latency requirements, and a need for instance-level explanations. Walk through how you'd compare linear regression, a single decision tree, Random Forest, and XGBoost.
Under what conditions would regularized linear regression actually beat tree-based models here, and when do ensembles clearly win?
Compare Random Forest and XGBoost in detail: training and inference cost, sensitivity to noisy features, overfitting risk, handling of missing values, and which hyperparameters most control bias versus variance.
When would Random Forest outperform XGBoost in a real production setting despite XGBoost's general reputation?
How would you generate fast and stable instance-level explanations for the model you pick, and how do you handle fairness checks and prediction calibration?
Design an experiment to confidently select the best model, covering cross-validation strategy, statistical testing across folds, and holdout confirmation.

“This was basically the whole interview wrapped in one prompt.”

View Post

Capital One·Data Scientist·Technical Phone Screen

Jun 2026

Capital One Data Scientist interview with a business case study on whether to launch a vegan burger product. Heavy on quantitative justification, which I wasn't fully expecting for a DS role.

Given revenue and cost figures for a vegan burger product, calculate the profit, profit margin, and break-even sales volume. Then, knowing that matching the standard burger's profit margin requires selling 60-70% more units, make a quantitative and qualitative case for or against launching the product. Also identify future trends that could make the vegan burger market more attractive.

“Took me a minute to organize this because it's really three questions stapled together.”

View Post

Capital One·Data Scientist·Technical Phone Screen

Jun 2026

Capital One data scientist interview with a case question that was more math-heavy than I expected. One question, but it had a lot of moving parts and I definitely fumbled the edge case interpretation.

A restaurant is considering adding a Vegan burger to its menu. The Regular burger earns $3 margin per unit, the Vegan earns $2. There are fixed costs F for launching the Vegan option (staff training, supplier fees), and a cannibalization rate c representing the share of Vegan sales that come from customers who would have otherwise bought a Regular. Derive the minimum Vegan unit sales needed to break even on profit, interpret what happens when c equals 0 or 1, and explain how you'd estimate c in practice.

“I got the algebra right but stumbled explaining the edge cases out loud.”

View Post

Capital One·Software Engineer·Technical Phone Screen

Jun 2026

Capital One SWE coding question, pretty focused on string manipulation and prefix logic. Nothing too wild but the edge cases will get you if you're not careful.

Given a list of words, count the number of pairs (i, j) where one word equals the other, or one word is a prefix of the other.

“My first instinct was brute force, just check every pair, which technically works but felt embarrassing to say out loud.”

View Post

Capital One·Software Engineer·Onsite - Product Sense / Strategy

Jun 2026

Capital One data engineer interview that leaned surprisingly heavy into business strategy territory. The main case was about whether to launch a credit limit increase program, which felt more like a product or finance role question than anything engineering-related. Wasn't totally prepared for the P&L framing.

You're supporting a consumer credit team considering a credit limit increase program for customers with 6 to 12 months of tenure and mid-range credit scores. Would you recommend launching it, and how would you size and target it?

“This one threw me more than I expected.”

View Post

Capital One·Data Scientist·Onsite - Cross-functional / Panel

Jun 2026

Capital One data scientist loop with a heavy focus on responsible AI and ethics around a face recognition use case at point of sale. Every question was a multi-part beast and the stakeholder pushback scenario at the end genuinely stressed me out.

What are the top risks of adding face recognition to verify cardholders at point of sale, how would you prioritize them, and who should own each one?
Draft a decision memo for this face recognition initiative covering the business purpose, relevant legal and regulatory considerations, a privacy impact assessment process, data minimization, retention schedule, and deletion workflows.
Define the go/no-go criteria and ongoing monitoring plan for this system, including accuracy thresholds broken down by demographic group, false match rate limits, human escalation triggers, adversarial testing, a rollback plan, and incident response timelines.
What less invasive alternatives could achieve the same fraud reduction or cardholder verification goal, and which one would you recommend and why?
A VP is pushing to launch despite unresolved fairness concerns. How do you push back, get stakeholders aligned, and propose a time-limited pilot with guardrails that could still result in a hard stop?

“I went straight to bias and disparate impact because that felt most obvious for a fintech context, then worked outward to privacy, consent, spoofing, data retention, and model governance.”

View Post