← Coinbase Interview Insights

Coinbase·Machine Learning Engineer·Technical Phone Screen·Intermediate

Intermediate
May 2026Remote

Summary

Coinbase ML engineer interview was a one-hour live coding session in a Jupyter notebook, basically build a classifier from scratch while someone watches you and asks questions. Not a whiteboard thing, actual messy data you have to wrangle yourself. Felt more like a working session than a traditional interview which I kind of appreciated.

Questions Asked (3)

Q1

Given a messy dataset in a Jupyter notebook, build a classification model end-to-end including light EDA, feature selection, model choice, fitting, and evaluation.

Technical Trade-offsData ModelingProduct Analytics & Metrics
Author's notes

The 'messy' part is real.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by framing the problem and defining success metrics, then walk through a structured pipeline: quick EDA to understand data quality and distributions, feature engineering and selection, model selection with rationale, training with validation, and evaluation using appropriate metrics. Emphasize trade-offs and decisions at each step, and mention how you would productionize or iterate.

Pro tip: Always start with a baseline model (e.g., logistic regression or majority class) to set a performance floor, and use cross-validation to avoid overfitting; this shows you value robustness over flashy results.

1. Understand the Problem and Data

Clarify the business objective, target variable, and success metrics. Perform light EDA: check data types, missing values, distributions, and class balance.

2. Preprocess and Feature Engineering

Handle missing values, encode categorical variables, scale numerical features, and create relevant features. Use domain knowledge to guide feature creation.

3. Feature Selection

Apply filter, wrapper, or embedded methods to select important features, reducing dimensionality and improving model performance.

4. Model Selection and Training

Choose candidate models (e.g., logistic regression, random forest, gradient boosting) based on data size, interpretability, and performance. Train with cross-validation and hyperparameter tuning.

5. Evaluation and Iteration

Evaluate using appropriate metrics (e.g., precision, recall, F1, AUC-ROC) and confusion matrix. Analyze errors, iterate on features or models, and consider deployment constraints.

Key Points to Mention

  • Handling missing data and outliers appropriately (imputation, removal, or robust methods)
  • Addressing class imbalance (resampling, class weights, or appropriate metrics)
  • Using cross-validation for reliable performance estimation and hyperparameter tuning
  • Selecting evaluation metrics aligned with business goals (e.g., precision vs. recall trade-off)
  • Considering model interpretability and explainability, especially for regulated domains like finance
  • Planning for productionization: model serialization, monitoring, and retraining

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

Q2

Why did you choose this particular model over other options?

Technical Trade-offsAdaptability & Ambiguity
Author's notes

Justified logistic regression as a baseline.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Frame your answer as a structured trade-off analysis: start by stating the business problem and constraints, then compare the chosen model against alternatives on specific axes like performance, latency, cost, and maintainability. Conclude by explaining how you validated the choice and what you would do if conditions changed.

Pro tip: Quantify trade-offs with real numbers from your experience (e.g., 'XGBoost gave 2% lower AUC but 10x faster inference than the neural net') and acknowledge that the 'best' model depends on context—this shows you optimize for business impact, not just metrics.

1. Define the problem and constraints

Briefly describe the business objective (e.g., fraud detection, price prediction) and key constraints such as latency, throughput, interpretability, and budget. This sets the stage for why certain models were even considered.

2. List alternatives considered

Name 2-3 other models you evaluated (e.g., logistic regression, random forest, transformer) and why they were plausible candidates. This shows you did a thorough search, not just picked a favorite.

3. Compare on key axes

Systematically compare the models on performance metrics (AUC, F1), inference speed, training cost, scalability, and maintainability. Use concrete numbers or relative comparisons to make the trade-offs tangible.

4. Explain the decision and validation

State why the chosen model won (e.g., best balance of accuracy and latency) and how you validated it (offline metrics, A/B test, shadow deployment). Mention any risks and mitigation.

5. Reflect on adaptability

Discuss what you learned and how you would revisit the choice if data volume, latency requirements, or business goals changed. This highlights your ability to handle ambiguity and evolving constraints.

Key Points to Mention

  • Business metrics vs. model metrics: tie model choice to business KPIs like fraud capture rate or customer experience.
  • Latency and throughput requirements: especially relevant for real-time systems like trading or fraud detection at Coinbase.
  • Interpretability and regulatory compliance: in fintech, being able to explain model decisions can be crucial.
  • Cost of training and inference: including infrastructure and engineering time.
  • Scalability and maintenance: how the model performs as data grows and how easy it is to retrain and monitor.
  • Validation strategy: offline evaluation, online A/B testing, and monitoring for drift.

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

Q3

Is accuracy the right evaluation metric here, given the class distribution in this dataset?

Data ModelingProduct Analytics & Metrics
Author's notes

Got caught flat-footed.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by acknowledging that accuracy can be misleading with imbalanced classes, then quantify the imbalance and discuss alternative metrics like precision, recall, F1, and AUC-ROC. Finally, tie the choice of metric to the business objective, such as minimizing false positives in fraud detection or maximizing recall for user safety.

Pro tip: At Coinbase, where fraud and security are critical, emphasize metrics that align with the cost of false positives vs. false negatives, and mention how you'd monitor the chosen metric in production.

1. Assess class distribution

Calculate the ratio of positive to negative classes and state whether it's imbalanced (e.g., 1:100). Explain how accuracy would be misleading, such as a naive model predicting the majority class achieving high accuracy.

2. Identify business costs

Discuss the relative costs of false positives and false negatives in the context of Coinbase's products (e.g., fraud detection, transaction monitoring). This determines whether precision or recall is more important.

3. Select appropriate metrics

Propose metrics like precision, recall, F1-score, AUC-ROC, or AUC-PR, and justify why they are better suited for imbalanced data. Mention that AUC-PR is often preferred when the positive class is rare.

4. Consider threshold and calibration

Explain that the decision threshold can be tuned to balance precision and recall based on business needs, and that probability calibration may be necessary for some metrics.

5. Validate with cross-validation

Emphasize using stratified cross-validation to ensure each fold preserves the class distribution, and to get reliable estimates of the chosen metrics.

Key Points to Mention

  • Class imbalance ratio and its impact on accuracy
  • Precision, recall, F1-score, and their trade-offs
  • AUC-ROC vs. AUC-PR for imbalanced datasets
  • Business context: cost of false positives vs. false negatives
  • Threshold tuning and probability calibration
  • Stratified cross-validation for reliable evaluation

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