Interview questions

Zoox Interview Questions and Process, Round by Round

April 30, 2026Updated October 19, 202611 min read
modern minimalist office

What each Zoox interview round covers, the coding, math and design questions candidates report, worked answers, and a two-week prep plan.

Zoox interview questions for software engineers come in four kinds: applied coding problems, some framed around vehicles; language depth in C++ or Python; system and class design; and a math and problem-solving round that usually involves no coding. Candidates describe a 30-minute recruiter call, one or two live coding screens, then a full-day onsite of five or six interviews, usually three to five weeks end to end. The math round is the one most candidates have not seen at other companies. It rewards talking through your reasoning out loud, not remembering formulas, and the rest of this guide shows what that looks like.

How the Zoox interview process runs

I found no official Zoox description of its hiring process, so everything below comes from prep sites that compile candidate reports and from candidates posting on Blind. Treat it as the likely shape, and ask your recruiter which rounds your team uses.

  1. Recruiter screen, about 30 minutes. Background, why autonomous vehicles, role fit, location, visa and timing.
  2. Technical phone screen. One or two live coding sessions of 60 to 90 minutes on a shared editor such as CoderPad, in C++ or Python. Expect questions on the language itself, not only a problem to solve.
  3. Onsite loop. A full day, often remote, of five or six interviews of 45 to 60 minutes: coding and debugging, low-level or object-oriented design, system design, math and science reasoning, and a behavioral interview with the hiring manager.

The mix varies by team. Design Gurus reports less math for infrastructure roles and more for robotics roles. TechPrep and Design Gurus both put the whole process at three to five weeks from first contact to offer.

The math and problem-solving round

This is the round behind searches for "Zoox math and problem solving interview questions". What candidates and Zoox employees have said about it on Blind:

  • One poster was asked to prep "high school algebra and geometry". A commenter who listed Zoox as their employer replied that the round "has been there since the day zoox started".
  • Another commenter who listed Zoox as their employer advised: "Make sure to explain your train of thought."
  • A 2023 candidate described about five questions covering "stuff up to basic trig", with an interviewer who "can give you hints".
  • A commenter compared it to "2010-era Google mathy word problems" and called it "more problem solving featuring math".
  • Prep guides describe a virtual whiteboard with no coding, covering geometry, trigonometry, physics and probability.

It is not only an engineering round. The 2023 thread was started by someone interviewing for a manufacturing role, and a software engineer asked about the same round later that year.

Three worked examples

The boat and the bowling ball. A Blind commenter shared this one: you are in a boat on a lake holding a bowling ball, and you throw it overboard. Does the lake level rise, fall or stay the same?

What it tests: whether you reach for the governing principle instead of guessing. A strong answer: while the ball is in the boat, the boat sits deeper and displaces extra water equal to the ball's weight. On the lake bottom, the ball displaces only its own volume. A bowling ball is denser than water, so the volume of water that weighs as much as the ball is larger than the ball itself. Less water is displaced after the throw, so the level falls. Say the assumption out loud (the ball sinks), then the principle (Archimedes), then the comparison. That order is what the interviewer is listening for.

A number check without a calculator. On another Blind thread, a commenter posted this as a joke (they said they had never interviewed at Zoox, so it is not a reported question), but it works as a practice problem for this kind of reasoning: is 3987¹² + 4365¹² = 4472¹² true?

What it tests: whether you can find a cheap invariant instead of brute-forcing. A strong answer: check divisibility by 3. The digits of 3987 sum to 27 and the digits of 4365 sum to 18, so both are multiples of 3, and so is the left side. The digits of 4472 sum to 17, so 4472 is not a multiple of 3, and neither is any power of it. The two sides cannot be equal. If you first try last digits, you will find both sides end in 6. Saying that check failed, and why you are switching to another one, is the kind of narration that earns credit.

A braking-distance estimate. This is a practice prompt, not a reported question, in the physics style the prep guides describe: a vehicle travels at 20 m/s and brakes at a steady 5 m/s². How far does it travel before stopping, and how much does a 100-millisecond delay before braking add?

A strong answer: stopping distance is v²/2a, so 400/10 = 40 metres. During a 100 ms delay the vehicle covers 20 × 0.1 = 2 metres at full speed, so 42 metres in total. Then say what the number means. Latency costs distance linearly with speed, and braking distance grows with the square of speed. That one sentence connects the maths to the product. TechPrep makes a similar point about system design, where it advises tying design choices to the physical system, such as how latency affects braking.

Coding questions: phone screen and onsite

TechPrep lists these as reported coding problems, several of them framed around driving or vehicles:

1. Simulate a lane change. Tests whether you can model state cleanly. Clarify first: discrete time steps or continuous? What counts as a safe gap? Then define a vehicle (lane, position, speed), write a function that checks the gap ahead and behind in the target lane, and step the simulation. Clean state and edge cases (the car at the end of the lane, two cars choosing the same gap) matter more than clever code.

2. Route across a city grid with roadblocks. Tests graph search choice. Say why you pick each tool: breadth-first search when every move costs the same, Dijkstra when edges carry weights, A* with a Manhattan-distance heuristic when you want to search fewer cells on a four-direction grid. Handle "no route exists" explicitly.

3. Merge overlapping intervals. Sort by start time, then walk the list, extending the current interval or starting a new one. O(n log n) for the sort. Edge cases: empty input, intervals that touch at a single point, one interval that swallows the rest.

4. Sliding window maximum. The deque version runs in O(n): keep indices in a deque with values decreasing, drop indices that fall out of the window from the front, drop smaller values from the back. Explain why each index enters and leaves once.

5. Best time to buy and sell a stock. One pass, tracking the lowest price so far and the best profit so far. Be ready for the multi-transaction variants, where dynamic programming is the natural next step.

6. Language internals. Reported examples include how C++ dispatches virtual functions through a vtable, pointer versus reference behaviour, and how Python handles memory for large datasets. For the pointer question, cover the basics: a reference must be bound when it is created, cannot be reseated and is never null in valid code, while a pointer can be all three. Pick the language you know best and know it well. Design Gurus notes that rejections are "often about depth, not breadth".

7. Debug someone else's code. TechPrep reports candidates being handed "a two-page code snippet" to fix or speed up under time pressure. Read it out loud, state what you think each block is meant to do, then test that against the code. Design Gurus says this round "rewards a spoken method, not silent staring".

Design rounds

System design: a versioning system for vehicle telemetry. This is the domain-flavoured prompt TechPrep reports. Start from the constraints, not boxes: how much data each vehicle produces, what must arrive in near real time and what can wait for an upload at the depot, and who reads it (engineers replaying a drive, teams rebuilding a dataset months later). A defensible outline treats raw logs as immutable chunks, keeps a catalogue that maps each dataset version to the exact chunks it contains, and makes any past dataset reproducible. Then name the tradeoffs you are making on storage cost, query speed and how stale a version can be.

Low-level design: a parking lot or a concurrent booking system. These are class-design exercises. Name the objects, their responsibilities and how they talk to each other, then defend the structure when the interviewer changes a requirement. For the booking version, concurrency is the real question: what stops two people taking the same seat?

Behavioral interview with the hiring manager

Prep guides describe this round as being about past projects, cross-team work, technical ownership and handling ambiguity. Prepare stories for these questions:

Why Zoox? A strong answer is specific to what Zoox builds. It designed its robotaxi from scratch rather than converting an existing car: no steering wheel, four seats facing each other, and a symmetrical body with four-wheel steering that lets it drive in both directions. In September 2025 it opened free public rides in Las Vegas; at that point it had not yet received approval to charge fares. Check Zoox's site for current service status. Pick the detail that connects to your own work, for example designing a system without a human fallback, and say why it interests you.

Tell me about a project you owned end to end. The hiring manager wants the decisions you made, the options you rejected and what you would change now. Keep the numbers consistent with anything you said earlier in the day. Design Gurus notes that interviewers compare notes.

Tell me about a time you made progress without clear requirements. Show how you defined the problem yourself, what you assumed, and how early you checked those assumptions.

Tell me about working across teams on something you did not control. Explain the other team's position fairly and what you gave up to reach agreement.

A two-week preparation plan

Days 1 to 4: coding. Work interval, sliding-window, graph-search and dynamic-programming problems in your main language, out loud, with a timer.

Days 5 to 6: language depth. Review how your language handles memory, references, virtual dispatch or object lifetimes. Write a few short answers you could give in two minutes.

Days 7 to 9: maths. Refresh algebra, basic trigonometry, units, kinematics and simple probability. Practise word problems out loud: assumptions, then principle, then calculation, then what the answer means.

Days 10 to 11: design. One data-heavy system design and one class design, each explained to a friend who keeps changing the requirements.

Days 12 to 14: stories and a mock day. Write four or five STAR stories, then run a mock day of back-to-back rounds to practise staying sharp late in the loop.

If you use Verve AI while preparing, its Knowledge Base takes your resume, the job description and STAR stories you write, and the Interview Copilot draws on the documents you switch on when it suggests answers. A practice session with a friend playing the hiring manager is a good place to find out which stories hold up under follow-up questions.

For related preparation, see our guides to Waymo interview questions and RTOS interview questions.

Sources

  • Recruiter screen length, phone screen format (60 to 90 minutes, CoderPad, C++ or Python), onsite of five to six 45 to 60 minute rounds, three to five week timeline, reported coding problems (lane change, grid routing with roadblocks, merge intervals, sliding window maximum, stock trading DP), telemetry data versioning design prompt, parking lot and concurrent booking design, C++ vtables, pointer vs reference, Python memory, two-page debugging snippet, math round on a virtual whiteboard (geometry, trigonometry, physics, probability), behavioral round focus: https://www.techprep.app/blog/zoox-interview-process
  • Recruiter screen topics (location, visa, timing), languages by area, round list, less math for infrastructure and more for robotics, three to five weeks, "rewards a spoken method", interviewers compare notes, "often about depth, not breadth": https://www.designgurus.io/answers/detail/what-is-the-zoox-interview-process-like-round-by-round
  • Math and problem solving round on a manufacturing-role onsite, about five questions, "stuff up to basic trig", hints, "2010-era Google mathy word problems", the boat and bowling ball question: https://www.teamblind.com/post/zoox-math-and-problem-solving-round-wwocca4d
  • "High school algebra and geometry", round "has been there since the day zoox started", "Make sure to explain your train of thought", the 3987¹² + 4365¹² problem posted as a joke by a commenter who had not interviewed at Zoox: https://www.teamblind.com/post/zoox-math-interview-standard-mklzko7d
  • A software engineer candidate asking about the Math/Problem Solving round (Nov 2023): https://www.teamblind.com/post/upcoming-zoox-interview-seeking-advice-for-mathproblem-solving-round-oasjixwf
  • Vehicle design (no steering wheel, four face-to-face seats, four-wheel steering, bidirectional travel): https://techcrunch.com/2021/06/22/how-amazon-owned-zoox-designed-its-self-driving-vehicles-to-prevent-crashes-and-protect-if-they-do/
  • Free public rides in Las Vegas from September 2025, pending approval to charge fares: https://www.fortune.com/2025/09/10/amazon-zoox-launches-limited-public-service-las-vegas
  • Knowledge Base accepts uploaded documents and written STAR stories; enabled documents shape answers: https://docs.vervecopilot.com/features/knowledge-base
  • Interview Copilot listens and shows suggestions only the user sees: https://docs.vervecopilot.com/features/interview-copilot
R

Ryan

CEO of Verve AI

Related reads

Explore Related Interview Guides

Why C++ Destructor Virtual Might Be The Most Crucial Concept For Object-oriented Success
August 6, 2025Interview prep guide

Why C++ Destructor Virtual Might Be The Most Crucial Concept For Object-oriented Success

Get insights on c++ destructor virtual with proven strategies and expert tips.

Read guide
Why C++ Projects Might Be Your Ultimate Edge In Any Professional Interview
August 8, 2025Interview prep guide

Why C++ Projects Might Be Your Ultimate Edge In Any Professional Interview

Get insights on c++ projects with proven strategies and expert tips.

Read guide
Why Can't I Find A Job Is More Than Just Bad Luck What Are You Missing
September 4, 2025Interview prep guide

Why Can't I Find A Job Is More Than Just Bad Luck What Are You Missing

Get insights on why can't i find a job with proven strategies and expert tips.

Read guide
Why Can't You Just Overload Functions In C: Mastering The C Overload Function Nuances For Interviews
August 15, 2025Interview prep guide

Why Can't You Just Overload Functions In C: Mastering The C Overload Function Nuances For Interviews

Get insights on c overload function with proven strategies and expert tips.

Read guide
Why Choosing The Right Acted Synonym Might Be Your Secret Interview Weapon
July 4, 2025Interview prep guide

Why Choosing The Right Acted Synonym Might Be Your Secret Interview Weapon

Get insights on acted synonym with proven strategies and expert tips.

Read guide
Why Copy Of Object Java Might Be Your Most Tricky Interview Question
August 6, 2025Interview prep guide

Why Copy Of Object Java Might Be Your Most Tricky Interview Question

Get insights on copy of object java with proven strategies and expert tips.

Read guide
Why Csharp Add To Array Is More Complex Than It Looks For Developers
August 5, 2025Interview prep guide

Why Csharp Add To Array Is More Complex Than It Looks For Developers

Get insights on csharp add to array with proven strategies and expert tips.

Read guide
Why Csharp Array Length Might Be The Most Overlooked Interview Detail
August 6, 2025Interview prep guide

Why Csharp Array Length Might Be The Most Overlooked Interview Detail

Get insights on csharp array length with proven strategies and expert tips.

Read guide
Why Describing What You Worked On Might Be Your Most Powerful Interview Tool
October 10, 2025Interview prep guide

Why Describing What You Worked On Might Be Your Most Powerful Interview Tool

Get insights on worked on with proven strategies and expert tips.

Read guide

Walk into every interview prepared and confident

Get Started For Free

Available on Mac, Windows, browser and mobile