Waymo Interview Questions
Reconstructed from 44 verified candidate reports across 28 questions. Aug 2025 – Jul 2026.
This page is a live view of every Waymointerview question AceOffer has indexed — pulled from real candidate reports, not invented from job descriptions or one founder’s memory. Every question shows how many times it’s been reported and when it was last seen. The catalog gets a refresh pass every month.
Key facts
- •28 distinct Waymo interview questions indexed
- •44 candidate reports across the catalog
- •Most reported: Shortest Path on Chess Board (BFS) — 3× (last seen April 2026)
- •Reports span Aug 2025 – Jul 2026
- •Refreshed monthly · last updated July 2026
Browse Waymo interviews by topic
The Waymo loop, from candidate reports
Waymo's loop is recruiter contact → a real technical phone screen (60 min live coding — BFS/graph problems and pure-Python data processing are the recurring themes, often with 'no pandas' as an explicit constraint) → a 3–5 round virtual onsite mixing coding, system design, a technical deep dive, and a hiring-manager behavioral. The timeline is fast: candidates report a median of 3 days from phone screen to onsite and about a week from onsite to outcome. System design rounds carry a strong autonomous-vehicle flavor — evaluation systems for self-driving models, simulation logging, ML inference at 100M-DAU scale — and interviewers probe domain specifics like the sim-to-real gap and compute-constrained simulation, so a generic template gets picked apart.
What gets asked, by round
Counts reflect distinct questions per round, not number of times asked. Frequencies on individual question cards show how many candidates reported getting that specific question.
60 minute design rounds. Interviewers push hard on the specific dimension their team cares about (storage at scale, real-time fan-out, multi-tenancy).
Hiring-manager conversation: project deep-dives with STAR structure, quantified outcomes, and role fit. Reported as the differentiator in otherwise-clean loops.
A real 60-minute live coding round, not a formality — graph/BFS problems and pure-Python data processing recur. You may be asked to write your own main() and test cases.
60–75 minute live coding rounds. Multiple sub-problems progressing in difficulty. Test harness usually provided.
Walk the interviewer through a past project end-to-end. Expect to defend technical choices and trace decisions to outcomes.
Most reported Waymo questions
Sorted by candidate-report frequency. These are the questions that have recurred most across the loops we’ve indexed.
| Question | Round | Reported | Last seen |
|---|---|---|---|
| Shortest Path on Chess Board (BFS) | Phone Screen | 3× | April 2026 |
| CSV Data Processing and Formatting | Phone Screen | 2× | April 2026 |
| Run Length Encoding Problems | Phone Screen | 2× | March 2026 |
| K-Means Clustering Implementation | Phone Screen | 2× | October 2025 |
| Custom String Sort by Character Order | Coding | 2× | March 2026 |
| Debug Coding (numpy/tensor/distributed) | Coding | 2× | June 2026 |
| Graph Search / BFS / DFS Coding | Coding | 2× | April 2026 |
| Largest Rectangle Area in Histogram (Variant) | Coding | 2× | April 2026 |
| Binary Search Optimization Problem | Coding | 2× | April 2026 |
| ETA / Shortest Path Calculation on Map | Phone Screen | 2× | May 2026 |
Want to see all 28? Browse the full Waymo catalog →
Read two Waymo questions free
Full problem statements, candidate-reported follow-ups, and walkthroughs. No signup needed.
The most-reported Waymo phone screen: BFS shortest path on a fixed board, with follow-ups extending to blocked cells and an infinite board. You're expected to write your own main() and test cases.
Design an evaluation system for self-driving models — human-in-the-loop plus LLM-based auto-eval. Interviewers probe the sim-to-real gap, running thousands of simulations with no compute budget, and live-demo bug handling.
Find planted bugs in numpy/tensor code: array aliasing in Matrix.zeros, omitted axis parameters, silent type truncation. The reported loss reason is missing the aliasing class of bug entirely.
- •On ML/inference system design: run the back-of-envelope first (QPS → memory → bandwidth → bottleneck) before proposing any optimization — passers structured the numbers up front
- •On open-ended AV design prompts (evaluation systems): anchor the scope fast with a concrete real-world analogy (candidates cited Scale AI / Mercor) and split human-eval vs LLM-eval into distinct subsystems
- •When one round chains multiple design areas (inference serving, model efficiency, kernel-level), keep breadth across all of them rather than going deep on one and running out of time
- •In the hiring-manager round: crisp STAR structure with quantified outcomes and ownership language ('I decided,' 'I designed') — and treat it as a two-way conversation with real questions about the team's roadmap
- •Align behavioral examples with the team's domain (planning, autonomy, simulation) — candidates who did reported better traction
- •Applying a generic system-design template without adapting to self-driving specifics — interviewers pivot to sim-to-real realism and compute constraints, and candidates who can't follow lose the round
- •Not drawing any diagrams on open-ended design rounds, leaving the discussion unstructured
- •Burning the first ~15 minutes of a 45-minute design slot on experience discussion or over-clarification, leaving no time for the actual design
- •Missing numpy/tensor traps in the debug coding round — array aliasing (Matrix.zeros), omitted axis parameters, silent type truncation
- •Giving a team-level project narrative without separating personal contribution, or citing no quantitative results — reported as the differentiator in otherwise-clean loops
Get the full Waymo catalog
Every question. Every candidate-reported follow-up. The mistakes that sink people, and what passers do instead. Monthly refresh.