← Waymo Interview Insights

Waymo·Data Scientist·Technical Phone Screen·Senior

Senior
Jun 2026

Summary

Waymo data scientist interview with a pretty intense stats/ML evaluation question focused on rare-event detection. The whole thing was essentially one big case study about model comparison under class imbalance, which I was not fully prepared for.

Questions Asked (3)

Q1

When comparing two models for rare-event detection (like fraud or medical adverse events), which metrics would you use and why is accuracy or ROC-AUC not sufficient on its own?

Product Analytics & MetricsTechnical Trade-offs
Author's notes

I knew accuracy was wrong immediately because of class imbalance, said so right away.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by acknowledging the class imbalance and the need for metrics that focus on the minority class. Then discuss precision-recall trade-offs, ranking metrics like PR-AUC, and cost-sensitive evaluation, explaining why accuracy and ROC-AUC can be misleading in rare-event detection.

Pro tip: Emphasize that the choice of metric should align with the business cost of false positives vs. false negatives, and mention that in practice, you might use a combination of metrics and set a threshold based on the desired precision or recall.

1. Acknowledge the rarity of events

State that rare events lead to highly imbalanced datasets, making accuracy a poor metric because a model can achieve high accuracy by always predicting the majority class.

2. Explain limitations of ROC-AUC

ROC-AUC can be overly optimistic because it considers both true positive rate and false positive rate, but in imbalanced settings, the false positive rate can be low even with many false positives, and it doesn't reflect the precision-recall trade-off.

3. Introduce precision-recall metrics

Recommend using precision, recall, F1-score, and especially the Precision-Recall AUC (PR-AUC) because they focus on the positive (rare) class and are more informative when the negative class dominates.

4. Consider cost-sensitive and ranking metrics

Mention that depending on the application, you might use cost-sensitive metrics (e.g., expected cost) or ranking metrics like lift, and that you should choose a threshold based on the desired balance of precision and recall.

5. Align with business objectives

Conclude that the metric should reflect the real-world costs of false positives and false negatives, and that you might combine multiple metrics for a comprehensive evaluation.

Key Points to Mention

  • Class imbalance and its effect on accuracy
  • ROC-AUC's insensitivity to class imbalance and its use of false positive rate
  • Precision-Recall curve and PR-AUC as better alternatives
  • F1-score as a balance between precision and recall
  • Cost-sensitive evaluation and threshold selection
  • Business context: cost of false positives vs. false negatives

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

Q2

How would you compare two models statistically when the number of positive examples is small? Walk through the distributions, confidence interval construction, and how you'd test for a significant difference.

A/B Testing & ExperimentationTechnical Trade-offs
Author's notes

This is where I got a little tangled.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by acknowledging that standard normal approximations fail with small positive counts, then propose exact or Bayesian methods. Walk through modeling each model's performance with a binomial or beta-binomial distribution, constructing intervals via Clopper-Pearson or Jeffreys, and testing differences with Fisher's exact test or a Bayesian posterior probability. Emphasize that the choice depends on assumptions and the decision context.

Pro tip: Mention that in safety-critical domains like autonomous driving, you should pre-register the analysis plan and consider Bayesian methods with informative priors to stabilize estimates, while also reporting frequentist results for robustness.

1. Clarify the metric and assumptions

Define what 'positive examples' mean (e.g., successful detections) and state assumptions like independence and identical distribution. Discuss whether the small sample is due to rarity or data collection limits.

2. Choose appropriate distributions

Model each model's positive count as Binomial(n, p) or use a Beta-Binomial if overdispersion is present. For very small counts, consider Poisson or negative binomial if events are rare.

3. Construct confidence intervals

Use exact methods like Clopper-Pearson for binomial proportions, or Bayesian credible intervals with a Beta prior (e.g., Jeffreys). Avoid Wald intervals due to poor coverage.

4. Test for significant difference

Apply Fisher's exact test for 2x2 tables, or a Bayesian test computing the posterior probability that one model is better. For paired data, use McNemar's test.

5. Interpret and communicate uncertainty

Report effect sizes with intervals, discuss practical significance, and acknowledge limitations. Recommend sequential testing or Bayesian decision rules if data is limited.

Key Points to Mention

  • Binomial and Beta-Binomial distributions for modeling counts
  • Clopper-Pearson and Jeffreys intervals for small samples
  • Fisher's exact test and its assumptions
  • Bayesian approaches with Beta priors and posterior probabilities
  • McNemar's test for paired comparisons
  • The impact of small samples on power and the need for pre-registration

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

Q3

If you only have a handful of aggregated numbers to work with (small sample, limited data), how do you make a responsible model selection decision? Consider threshold choice, calibration quality, and alert-budget constraints.

Technical Trade-offsAdaptability & Ambiguity
Author's notes

Liked this one more.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Acknowledge the data limitation upfront, then propose a principled approach that combines uncertainty quantification (e.g., Bayesian methods or bootstrapping) with domain constraints. Emphasize that with small samples, the goal shifts from maximizing accuracy to making robust, conservative decisions that respect alert budgets and safety requirements.

Pro tip: Show that you understand the operational context: at Waymo, false negatives can be catastrophic, so you might prioritize recall over precision and use a threshold that minimizes missed detections, even if it means more false alerts—but then manage the alert budget by tiering or deferring low-confidence cases.

1. Quantify uncertainty and set expectations

Use Bayesian credible intervals or bootstrap resampling to estimate the uncertainty around any metric. Communicate that with small samples, point estimates are unreliable and decisions should be based on conservative bounds.

2. Incorporate domain constraints and costs

Define the relative cost of false positives vs. false negatives and the hard limit on alert volume. This turns the problem into a constrained optimization: maximize detection under the alert budget.

3. Choose threshold and calibration method robustly

Prefer simple, well-calibrated models (e.g., logistic regression) and use cross-validation or holdout with uncertainty. For threshold selection, use cost-sensitive learning or optimize a metric like F-beta that weights recall higher.

4. Evaluate calibration quality with limited data

Assess calibration via reliability diagrams or metrics like Brier score, but acknowledge small-sample noise. Consider using isotonic regression or Platt scaling with regularization, and validate on held-out data if possible.

5. Decide and monitor with a fallback plan

Make a decision that is defensible given uncertainty, and set up a monitoring plan to collect more data and re-evaluate. If possible, use a human-in-the-loop or tiered alerting to mitigate risk.

Key Points to Mention

  • Bayesian methods for uncertainty quantification (e.g., credible intervals, posterior predictive checks)
  • Cost-sensitive learning and asymmetric misclassification costs
  • Alert budget as a constraint: precision-recall trade-off and threshold optimization
  • Calibration metrics (reliability diagrams, Brier score) and methods (Platt scaling, isotonic regression)
  • Cross-validation and bootstrapping for small samples
  • Domain-specific safety considerations (e.g., high recall for safety-critical systems)

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