Waymo Data Scientist interview questions

10 Data Scientists have shared notes from their Waymo interviews, with the questions they were asked in each round. That is 11% of the 87 Waymo interview notes on Screna AI.

10 Data Scientist interview notes

Waymo·Data Scientist·Technical Phone Screen

Jul 2026
Waymo data science interview with a meaty stats/ML problem centered on trajectory analysis. The whole session was basically one long case question with a follow-up baked in. Felt more like a research discussion than a traditional interview.
  • You have drone-captured 2D turning trajectories from vehicles at multiple intersections, some from Waymo autonomous vehicles and some from others. How would you statistically test whether Waymo's turning behavior is meaningfully different from the rest of the population? Walk through your feature choices, distributional assumptions, test selection, and how you'd report effect size.
  • Traffic volume and driving patterns vary across intersections. How do you prevent intersection-level confounding from distorting your Waymo vs. other vehicles comparison, and what modeling approach would you use?

“This is a deceptively open question.”

View Post

Waymo·Data Scientist·Technical Phone Screen

Jul 2026
Waymo data scientist interview with two pretty distinct sections: one on K-means from scratch and one on multi-agent trajectory prediction that got pretty deep into transformer architectures and training-inference mismatch. The ML modeling half felt more like a research discussion than a standard DS interview.
  • Explain what objective K-means optimizes and walk through how the alternating optimization procedure works.
  • Implement K-means (Lloyd's algorithm) from scratch, including centroid initialization, the assignment and update steps, and a stopping condition.
  • Describe and implement a better centroid initialization strategy than random initialization.
  • Propose an ML approach to predict the next 2 positions of a target vehicle given its past trajectory, nearby agent trajectories, and map context. Cover input representation, architecture, output parameterization, and loss/metrics.
  • Explain what multi-head attention is doing in the context of multi-agent trajectory prediction and why it's useful there.
  • During autoregressive decoding, the model sees ground-truth previous steps at training time but its own (potentially wrong) predictions at inference time. How would you fix this mismatch, and what are the tradeoffs of different approaches?

“Pretty standard setup question but I fumbled the formal objective a bit.”

View Post

Waymo·Data Scientist·Technical Phone Screen

Jun 2026
Waymo data scientist interview, SQL-heavy technical screen with four progressively harder questions all built around the same ride-hailing schema. The last two questions were genuinely tricky and I don't think I fully nailed either of them.
  • Given a users table and a rides table for a ride-hailing product, count the number of users who took zero rides during a specific 7-day window (July 1 through July 7, 2024, inclusive).
  • For each calendar month that had at least one ride, compute the total number of rides and the average ride rating rounded to two decimal places. Return results sorted by month descending.
  • Count the number of distinct users who have taken 0 or 1 total rides across their entire lifetime, grouped by city. Users who never took a ride must be included in the count.
  • For each user, rank the calendar months in which they had at least one ride chronologically (skipping months with no rides). Then compute the average ride rating across all rides that occurred in users' 1st and 3rd active months, grouped by city.

“Pretty straightforward once you remember to left join from users and filter on the ride side for nulls.”

View Post

Waymo·Data Scientist·Technical Phone Screen

Jun 2026
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.
  • 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?
  • 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.
  • 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.

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

View Post

Waymo·Data Scientist·Technical Phone Screen

Jun 2026
Waymo data scientist interview with a pretty gnarly statistics/sampling problem. One question, but it had enough layers to keep me busy for a while. Not your typical SQL or A/B test setup.
  • You run a restaurant with 10,000 reservations in a day. Each reservation has a reserved party size and an actual show-up count. You only observe 1,000 of these reservations, but the sampling is size-biased: larger reservations are proportionally more likely to be sampled. How would you estimate the total number of people who actually show up across all 10,000 reservations, and how would you quantify uncertainty around that estimate?

“The size-biased sampling part is what trips you up if you're not careful.”

View Post

Waymo·Data Scientist·Technical Phone Screen

Jun 2026
Waymo data scientist interview with a meaty product analytics case built around a ride-hailing feature called Smart Wait. The whole thing was basically one extended scenario with four parts, and the difficulty ramped up fast once you got past the surface-level agreement or disagreement question.
  • A PM says fleet efficiency improved because actual time to pickup decreased after the Smart Wait feature launched. Do you agree with that conclusion?
  • What alternative explanations could account for the decrease in time to pickup, particularly given that conversion dropped at the same time?
  • What metrics and analyses would you use to determine whether Smart Wait was actually a net positive for the product?
  • If you were running an experiment on this feature, what would you randomize, what would be your primary metric, and what guardrail metrics would you track?

“I said yes too quickly and then had to walk it back.”

View Post

Waymo·Data Scientist·Technical Phone Screen

May 2026
Waymo Data Scientist screen, pretty straightforward Python fundamentals. One coding question, nothing that required deep ML knowledge, which surprised me a bit.
  • Write a Python function that computes the arithmetic mean of a list of numbers, returning 0 for an empty list and without modifying the input.

“Seemed almost too easy so I second-guessed myself for a second.”

View Post

Waymo·Data Scientist·Technical Phone Screen

May 2026
Waymo data scientist interview with a pretty heavy computational geometry problem. Not what I expected going in, felt more like a physics/algorithms hybrid than anything data-sciency.
  • Given n cars on a 1D road, each with an initial position, velocity, constant acceleration, and a physical radius, determine whether any two cars collide for t >= 0. If a collision exists, return the earliest collision time and the pair involved.
  • Extend the collision detection to 2D: each car moves on a flat plane with a 2D position, velocity, and acceleration vector, plus a radius. Find the earliest collision time across all pairs.
  • The naive approach checks all pairs in O(n^2). How would you optimize collision detection for large n, say up to 10^5 cars?
  • How do you handle the edge case where two cars have identical trajectories and overlap for an entire time interval rather than a single point?

“The math itself isn't too bad once you set it up.”

View Post

Waymo·Data Scientist·Technical Phone Screen

Apr 2026
Waymo data scientist interview that leaned way harder into probability theory than I expected. The whole session was basically one extended stats problem with a simulation layer bolted on, which was fine until the confidence interval follow-up made me realize I hadn't thought carefully enough about the variance.
  • In a best-of-5 table tennis match where player A wins each game with fixed probability p, derive the probability that the match goes to exactly 5 games.
  • How would you write a Monte Carlo simulation to estimate the probability that a best-of-5 match lasts exactly 5 games?
  • Given M simulated matches, how would you construct a confidence interval for the estimated probability?

“This took me longer to set up than it should have.”

View Post

Waymo·Data Scientist·Technical Phone Screen

Jul 2024
Waymo data scientist interview that was pretty much a single extended case study on A/B testing for a routing algorithm. The whole thing revolved around experiment design and statistical validity, which I thought I knew well until the follow-ups made me second-guess myself.
  • Write Python code to filter ride data to San Francisco in July 2024, join it to experiment assignments, run a Welch two-sample t-test comparing treatment vs. control time-to-pickup, and return the p-value.
  • Given a returned p-value, how would you decide whether the result is statistically significant?
  • Is a statistically significant p-value conclusive proof that the new routing algorithm is better? If not, what are the main threats to validity?
  • Propose a stronger experiment design and analysis plan for this routing algorithm.

“Felt okay about the pandas part but I second-guessed myself on why Welch specifically versus a standard t-test.”

View Post