← Capital One Interview Insights

Capital One·Data Scientist·Onsite - System Design / Architecture·Senior

SeniorPrefer not to say
May 2026

Summary

Capital One system design round for a Data Scientist role, and this one was a beast. The whole session was basically a single massive question about building an on-device face recognition system at scale, and they kept pulling threads on every answer I gave.

Questions Asked (9)

Q1

Design an on-device face recognition system for mobile access control with 50 million monthly active users and intermittent connectivity. Should the system do verification or identification, and why?

System DesignTechnical Trade-offsProduct Strategy
Author's notes

I went with verification pretty quickly since identification at 50M scale on-device is basically impossible without a server.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the access control use case: for a single user unlocking their own device or account, verification (1:1) is the right choice because it is faster, more secure, and works offline. Then discuss how identification (1:N) would be needed for scenarios like identifying any user in a group, but that introduces scalability and privacy challenges that are impractical on-device for 50M users.

Pro tip: Emphasize that verification is not just a technical choice but a product and privacy decision: it minimizes on-device storage and computation, aligns with user expectations of personal device security, and avoids the legal and ethical risks of a large-scale biometric identification database.

1. Clarify the use case and constraints

Define the access control scenario: is it unlocking a personal device, accessing a building, or authorizing a transaction? Consider the 50M MAU and intermittent connectivity, which demand on-device processing and offline capability.

2. Compare verification vs. identification

Explain that verification (1:1) confirms a claimed identity, while identification (1:N) determines who the person is from a database. For personal access control, verification is typically sufficient and more efficient.

3. Evaluate technical trade-offs

Discuss on-device constraints: verification requires storing only one template per user, enabling fast, low-power matching. Identification would require storing and searching a large gallery, which is infeasible on-device for 50M users and raises privacy concerns.

4. Address connectivity and scalability

Highlight that verification works offline, ensuring access even without connectivity. For identification, intermittent connectivity would require syncing a massive database, which is impractical and introduces latency and security risks.

5. Conclude with a recommendation

Recommend verification for on-device access control, as it balances security, usability, and scalability. Mention that identification could be a server-side fallback for specific high-security scenarios, but not for the core on-device system.

Key Points to Mention

  • Verification is 1:1 and identification is 1:N; access control for personal devices is inherently a verification problem.
  • On-device verification requires storing only one biometric template per user, reducing storage and computation needs.
  • Identification would require a large on-device gallery, which is impractical for 50M users and raises privacy and legal issues.
  • Intermittent connectivity favors verification because it can operate fully offline without syncing a database.
  • Verification provides faster response times and lower power consumption, critical for mobile user experience.
  • Security and privacy: verification minimizes the risk of false matches and avoids creating a centralized biometric database.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q2

What model family and embedding dimension would you choose for this face recognition system, and what training objective would you use?

System DesignTechnical Trade-offsAlgorithms & Data Structures
Author's notes

Went with a MobileNet-style backbone and 128-dim embeddings.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the requirements: scale, latency, accuracy, and deployment constraints. Then propose a model family (e.g., ArcFace with a ResNet backbone) and embedding dimension (e.g., 512) justified by trade-offs. Finally, describe the training objective (e.g., ArcFace loss) and how it addresses the problem.

Pro tip: Mention that embedding dimension is a hyperparameter that should be tuned based on validation performance and computational budget, and that larger dimensions may not always yield better accuracy due to overfitting. Also, highlight the importance of using a margin-based softmax loss for discriminative embeddings.

1. Clarify Requirements

Ask about scale (number of identities), latency, accuracy targets, and deployment environment (cloud vs. edge). This ensures your choices align with business needs.

2. Choose Model Family

Recommend a state-of-the-art face recognition model like ArcFace, CosFace, or a Vision Transformer, explaining why it suits the requirements (e.g., high accuracy, robustness).

3. Select Embedding Dimension

Propose a dimension (e.g., 512) and justify it by discussing trade-offs: higher dimensions capture more detail but increase storage and computation; lower dimensions are efficient but may lose discriminative power.

4. Define Training Objective

Describe the loss function, such as ArcFace (additive angular margin loss), and explain how it enhances intra-class compactness and inter-class separability.

5. Discuss Evaluation and Iteration

Mention metrics (e.g., TAR@FAR, accuracy) and the need to validate choices, possibly with A/B testing or cross-validation, and iterate if needed.

Key Points to Mention

  • ArcFace or similar margin-based softmax loss for discriminative embeddings
  • Embedding dimension trade-offs: 128 vs. 256 vs. 512 vs. 1024
  • Backbone architecture: ResNet-100, MobileFaceNet for efficiency, or Vision Transformers
  • Training data requirements: large-scale datasets like MS1M or Glint360K
  • Evaluation metrics: TAR@FAR, ROC curves, and identification accuracy
  • Deployment considerations: model size, inference speed, and hardware constraints

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q3

How would you evaluate this face recognition model? Describe your protocol including ROC/DET curves and specific threshold targets.

Product Analytics & MetricsTechnical Trade-offs
Author's notes

Standard FAR/FRR tradeoff stuff, I talked through setting FAR at 0.001 with TPR above 0.98.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the business context and operational requirements, then outline a comprehensive evaluation protocol that includes ROC/DET curves, threshold selection based on cost-benefit analysis, and validation on diverse datasets. Emphasize how you would translate model performance into actionable metrics for stakeholders.

Pro tip: In fraud detection, the cost of false negatives (missed fraud) is often much higher than false positives, so choose thresholds that minimize expected cost rather than optimizing for accuracy. Also, consider using precision-recall curves when dealing with highly imbalanced data, as ROC can be overly optimistic.

1. Define Business Objectives and Constraints

Clarify the specific use case (e.g., fraud detection, customer verification) and the associated costs of false positives and false negatives. Determine regulatory and fairness constraints.

2. Select Evaluation Metrics and Curves

Choose appropriate metrics such as ROC-AUC, DET curves, and precision-recall curves. Explain how each provides insights into model performance across different thresholds.

3. Determine Threshold Targets

Use cost-benefit analysis to set thresholds that align with business goals. For example, set a threshold to achieve a desired false acceptance rate (FAR) or false rejection rate (FRR).

4. Validate on Diverse and Representative Data

Test the model on multiple datasets that reflect real-world variability, including different demographics, lighting conditions, and image qualities. Assess performance consistency and fairness.

5. Monitor and Iterate Post-Deployment

Implement ongoing monitoring for performance drift and bias. Establish a feedback loop to retrain and adjust thresholds as needed.

Key Points to Mention

  • ROC curve and AUC as a threshold-independent measure of discriminative ability
  • DET curve for visualizing trade-offs between false acceptance and false rejection rates
  • Precision-recall curves for imbalanced datasets
  • Threshold selection based on business costs (e.g., cost matrix)
  • Equal error rate (EER) as a summary metric
  • Fairness and bias evaluation across demographic groups

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q4

How would you handle liveness detection and anti-spoofing, including dealing with occlusions like masks and glasses?

System DesignTechnical Trade-offs
Author's notes

Talked about 2D texture analysis as baseline, IR for depth on supported hardware, and mentioned challenge-response as a fallback.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the business context and threat model, then propose a layered defense combining passive and active liveness detection. Discuss how to handle occlusions through robust model design, data augmentation, and fallback mechanisms, while balancing security, user experience, and regulatory constraints.

Pro tip: Emphasize that liveness detection is an adversarial problem, so you must continuously monitor and update models to counter new spoofing techniques. Also, highlight the importance of explainability and fairness to meet financial regulations and avoid bias against users with occlusions.

1. Clarify Requirements and Threat Model

Ask about the specific use case (e.g., account opening, high-value transactions), acceptable false accept/reject rates, and potential spoofing attacks (printed photos, replays, 3D masks).

2. Design a Multi-Layered Liveness Detection System

Combine passive (texture, depth, motion analysis) and active (challenge-response) methods. Use deep learning models like CNNs or transformers for feature extraction, and consider multi-modal fusion (RGB, depth, IR).

3. Address Occlusions Robustly

Train with augmented data including masks, glasses, and varying lighting. Use attention mechanisms or region-based processing to focus on visible facial areas, and implement fallback to alternative verification if occlusion is too severe.

4. Evaluate and Iterate with Metrics and Adversarial Testing

Define metrics like APCER, BPCER, and EER. Conduct red-team exercises to simulate spoofing attempts, and monitor performance in production to detect drift and new attack vectors.

5. Balance Security, UX, and Compliance

Optimize thresholds to minimize friction while meeting regulatory requirements (e.g., PSD2, KYC). Ensure fairness across demographics and provide clear user guidance for occluded scenarios.

Key Points to Mention

  • Passive vs. active liveness detection and their trade-offs
  • Use of multi-modal biometrics (e.g., depth, infrared) to improve robustness
  • Data augmentation and synthetic data to simulate occlusions
  • Adversarial attacks and defenses (e.g., GAN-generated spoofs)
  • Metrics: APCER, BPCER, EER, and their business implications
  • Regulatory and ethical considerations (fairness, privacy, explainability)

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q5

How would you ensure demographic fairness in the system, including threshold calibration across different user cohorts?

Product Analytics & MetricsTechnical Trade-offsAdaptability & Ambiguity
Author's notes

This one tripped me up more than I expected.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining fairness metrics and identifying protected cohorts, then describe a threshold calibration process that balances fairness and performance. Emphasize iterative monitoring and trade-off discussions with stakeholders to ensure alignment with business and regulatory goals.

Pro tip: Acknowledge that fairness is context-dependent and there's no one-size-fits-all solution; showing awareness of legal and ethical considerations while proposing a pragmatic, data-driven approach will set you apart.

1. Define fairness metrics and cohorts

Select appropriate fairness definitions (e.g., demographic parity, equal opportunity) and identify relevant user cohorts based on demographics and business context.

2. Measure baseline disparities

Evaluate model performance across cohorts using the chosen metrics to quantify any existing biases or disparities in outcomes.

3. Calibrate thresholds per cohort

Adjust decision thresholds for each cohort to achieve fairness goals, using techniques like group-specific thresholds or post-processing calibration.

4. Validate and monitor

Continuously monitor fairness metrics and model performance, with regular audits and feedback loops to detect and mitigate emerging biases.

5. Communicate trade-offs

Discuss trade-offs between fairness, accuracy, and business objectives with stakeholders to ensure transparent decision-making and alignment.

Key Points to Mention

  • Fairness definitions (demographic parity, equal opportunity, etc.)
  • Cohort analysis and segmentation
  • Threshold calibration methods (group-specific thresholds, post-processing)
  • Trade-offs between fairness and model performance
  • Regulatory and ethical considerations (e.g., ECOA, GDPR)
  • Continuous monitoring and auditing for fairness

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q6

What privacy and security measures would you implement, covering on-device template storage, differential privacy, template protection, and replay attack prevention?

System DesignTechnical Trade-offs
Author's notes

Covered on-device encrypted storage, no raw biometric transmission, and fuzzy commitment schemes for template protection.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by framing the problem around the CIA triad and regulatory requirements, then systematically address each area: on-device template storage, differential privacy, template protection, and replay attack prevention. For each, explain the mechanism, trade-offs (e.g., security vs. utility, latency vs. privacy), and how you would validate effectiveness. Conclude by emphasizing a layered defense strategy and continuous monitoring.

Pro tip: In regulated industries like banking, always tie technical choices to compliance frameworks (e.g., GDPR, CCPA, PSD2) and quantify trade-offs—e.g., how differential privacy's epsilon affects model accuracy—to show business awareness.

1. Clarify requirements and threat model

Ask clarifying questions about data types, regulatory constraints, and potential adversaries. Define what 'privacy' and 'security' mean in this context (e.g., biometric templates, user behavior data).

2. Design on-device template storage

Propose secure storage using hardware-backed keystores (e.g., Secure Enclave, TEE) and encryption at rest. Discuss trade-offs between local processing and cloud sync, emphasizing minimization of data leaving the device.

3. Apply differential privacy

Explain how to add calibrated noise to aggregated data or model updates to protect individual privacy. Discuss choosing epsilon based on utility needs and techniques like local vs. global DP.

4. Implement template protection

Describe irreversible transformations such as fuzzy hashing, homomorphic encryption, or secure multi-party computation to protect templates even if breached. Highlight trade-offs in matching accuracy and computational cost.

5. Prevent replay attacks

Use nonces, timestamps, and challenge-response protocols to ensure freshness. Combine with liveness detection for biometrics and rate limiting to mitigate replay attempts.

Key Points to Mention

  • Hardware-backed security (TEE, Secure Enclave) for on-device storage
  • Differential privacy: epsilon-delta trade-offs, local vs. global DP
  • Template protection: irreversible transforms (e.g., fuzzy hashing, homomorphic encryption)
  • Replay attack prevention: nonces, timestamps, challenge-response, liveness detection
  • Compliance with regulations (GDPR, CCPA, PSD2) and data minimization
  • Layered defense: encryption, access controls, auditing, and continuous monitoring

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q7

What are your latency, memory, and battery constraints for this system on a mid-tier mobile device, and how would you meet them?

System DesignTechnical Trade-offs
Author's notes

p95 under 150ms, model under 50MB, I said quantization and pruning get you there on most modern mid-tier chips.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the specific mobile use case and data pipeline, then propose realistic constraints for a mid-tier device (e.g., <100ms inference latency, <50MB memory, <2% battery per hour) and justify them with user experience and business impact. Finally, outline a layered strategy combining model optimization, on-device/cloud trade-offs, and continuous monitoring to meet those constraints.

Pro tip: Anchor your constraints in a concrete user scenario (e.g., real-time fraud detection during checkout) and quantify the cost of failure—this shows you understand that constraints are business-driven, not just technical.

1. Clarify the use case and device profile

Ask about the specific mobile application, expected user interactions, and target device specifications (CPU, RAM, battery) to ground your constraints in reality.

2. Propose and justify constraints

State concrete numbers for latency (e.g., <100ms for real-time), memory (e.g., <50MB), and battery (e.g., <2% per hour) and explain how they derive from user experience and business requirements.

3. Outline optimization techniques

Describe model compression (quantization, pruning, knowledge distillation), efficient architectures (MobileNets, TinyML), and on-device vs. cloud inference trade-offs to meet the constraints.

4. Address trade-offs and monitoring

Discuss how you balance accuracy vs. efficiency, and propose a monitoring plan to track latency, memory, and battery in production and adapt as needed.

Key Points to Mention

  • Quantization and pruning to reduce model size and inference time
  • On-device inference vs. cloud offloading trade-offs (latency, privacy, connectivity)
  • Battery impact of continuous inference and strategies like batching or event-driven triggers
  • Memory footprint management: model size, runtime memory, and caching
  • Use of hardware acceleration (GPU, NPU) on mid-tier devices
  • Continuous monitoring and A/B testing to validate constraints in real-world conditions

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q8

How would you monitor the deployed system for model drift and manage periodic re-enrollment of users?

Product Analytics & MetricsRoot Cause Analysis
Author's notes

Talked about tracking match score distributions over time as a drift signal and flagging users whose scores degrade past a threshold for re-enrollment prompts.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining model drift and its types (data drift, concept drift), then outline a monitoring framework using statistical tests and business metrics. Next, describe a re-enrollment strategy that balances model performance with user experience, including triggers and communication plans. Finally, tie it back to Capital One's regulatory environment and customer-centric values.

Pro tip: Emphasize the importance of setting up automated alerts with thresholds based on business impact, not just statistical significance, and always have a rollback plan. Mention that re-enrollment should be seamless for users, with clear communication to maintain trust.

1. Define drift and monitoring metrics

Identify what constitutes drift for your model (e.g., feature distribution shifts, prediction drift, performance degradation) and select appropriate metrics (PSI, KL divergence, accuracy, F1).

2. Implement monitoring pipeline

Set up automated data collection, compute drift metrics on a schedule, and create dashboards with alerts for when thresholds are breached.

3. Root cause analysis and impact assessment

When drift is detected, investigate causes (e.g., seasonality, new user behavior, data pipeline issues) and quantify impact on business KPIs.

4. Design re-enrollment strategy

Determine triggers for re-enrollment (e.g., drift severity, performance drop), frequency, and method (e.g., gradual rollout, A/B test) while minimizing user friction.

5. Execute and communicate

Roll out re-enrollment in phases, monitor outcomes, and communicate changes to users and stakeholders, ensuring compliance with regulations.

Key Points to Mention

  • Types of drift: data drift, concept drift, label drift
  • Statistical tests: PSI, KL divergence, KS test
  • Business metrics: conversion rate, customer satisfaction, revenue impact
  • Automated monitoring tools: Evidently AI, WhyLabs, custom dashboards
  • Re-enrollment triggers: performance thresholds, time-based, event-based
  • User communication and consent, especially in regulated industries

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q9

How would you safely A/B test a face recognition system like this, including shadow mode rollout and guardrails?

A/B Testing & ExperimentationSystem Design
Author's notes

Shadow mode was my main answer, run the new model in parallel without acting on its decisions, compare outcomes against the production model before any traffic switch.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by emphasizing safety and compliance, then outline a phased rollout: offline evaluation, shadow mode, canary testing, and finally A/B test with strict guardrails. Focus on defining clear success metrics, monitoring for bias and fairness, and having rollback plans.

Pro tip: In regulated industries like finance, always involve legal and compliance early, and document everything—especially how you handle PII and model explainability. This shows maturity beyond just technical execution.

1. Define Objectives and Metrics

Clearly state the goal of the A/B test (e.g., improve accuracy, reduce false positives) and define primary and guardrail metrics (e.g., false positive rate, demographic parity, latency).

2. Offline Evaluation and Shadow Mode

First, evaluate the new model offline on historical data. Then, deploy in shadow mode where it runs in parallel with the current system but its predictions are not used, to compare performance without risk.

3. Canary Rollout and A/B Test Design

Gradually roll out to a small percentage of traffic (canary) to catch issues. Then, design a randomized controlled experiment with proper sample size and randomization unit (e.g., user ID) to compare treatment and control.

4. Monitor Guardrails and Fairness

Continuously monitor guardrail metrics for degradation, including fairness across demographic groups. Set up alerts and automated rollback if thresholds are breached.

5. Analyze Results and Decide

After sufficient data, analyze results with statistical tests, check for novelty effects, and decide whether to fully launch, iterate, or roll back based on both primary and guardrail metrics.

Key Points to Mention

  • Shadow mode: run new model in parallel without affecting users to compare predictions safely.
  • Guardrail metrics: define thresholds for false positives, false negatives, bias, and latency; monitor and alert.
  • Fairness and bias testing: evaluate performance across demographic groups to avoid discrimination.
  • Regulatory compliance: adhere to GDPR, CCPA, and financial regulations; involve legal and compliance teams.
  • Rollback plan: have a clear process to revert to the old model if issues arise.
  • Sample size and power: ensure the experiment has enough data to detect meaningful differences.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.