Zoox System Design Interview Questions
- Tell the class-design coding round apart from the classic system design round
- Know which levels and tracks get a system design round, and what replaces it on autonomy, ML and data teams
- Know the fleet-shaped 2026 prompts and how to work one end to end
When Zoox candidates say "the design round," they mean one of two different things. The first is the object-oriented coding problem that sits in nearly every software screen: you model a small real-world scenario in classes and the code has to run. The second is a classic system design round on a whiteboard, with services, storage and data flowing between them. It's common, but not every loop has one, and it depends on your level and your team.
The 2026 picture is clear enough to plan around. Asked about Zoox and Waymo loops in October 2026, a Zoox engineer said both companies' system design rounds are "domain focused, won't be like design a Uber/Slack/etc." The prompts reported this year bear that out. They're about the fleet: where the robotaxis are, which ones are nearby, and what a ride costs.
Two different rounds both get called design
The class-design coding round. It's the screen for most software tracks, sometimes an onsite slot too, and it has kept the same shape for years. You get 10 to 15 minutes of quick language trivia, then a practical class-design problem in CoderPad that has to pass the given input. The newest named one is roller coaster scoring (October 2026, onsite day one), and airline ticket pricing is the long-running one (2021 to 2025). Recruiters and candidates file it under "system design," "software design" or "OOD," which is why reports blur together. It's graded as running code, and clean classes count for more than a quick hard-coded answer. Both problems are worked in Class Design: Roller Coaster Scoring & Airline Pricing.
The classic system design round. This one is a whiteboard conversation about a distributed system. In 2026 it shows up in two places. Some candidates get it inside the technical screen: a nearby-search prompt in May and a live-map prompt in August both came up at that stage. Others get it in the onsite, next to coding, the math and problem solving round and the hiring manager. A Senior/Staff loop in August 2026 had two of them.
It's backend-heavy, even for frontend engineers. A frontend-focused candidate who failed it in May 2026 wrote: "their system design is more backend-heavy, so if you're a frontend-focused engineer like me, be prepared to go deeper on backend architecture than you might expect." Another frontend-leaning candidate asked about it in July 2026, and a Zoox engineer told them to work through general system design guides and high-level architecture.
Does Zoox have a system design interview at your level?
Usually yes for generalist software roles, though how many rounds and what kind depend on level and track. Autonomy, ML, data and safety research teams swap in their own kind of design. Ask your recruiter which rounds your loop has, because team-to-team variation is real at Zoox.
| Level or track | System design round? | What it looks like |
|---|---|---|
| L3 and L4 software engineer | Often one, team dependent | One round in many onsites, and in 2026 sometimes inside the technical screen. One 2025 loop had a domain round instead. |
| L5 Senior | Usually one | A classic round in most senior loops. Two senior loops in 2024 had none, one of them on an autonomy team. |
| Senior/Staff and Staff | Possibly two | An August 2026 Senior/Staff loop had two system design rounds. |
| Engineering manager | One, team specific | A 2024 manager said to expect a problem the team is working on right now. |
| Autonomy (motion planning, perception) | Domain rounds instead | Design shows up as a motion-planning class problem in C++ and in domain discussions. |
| ML engineer | ML system design | Part of 2025 senior and mid-level loops, with prompts set by the team. |
| Data engineer | Data design | A data model plus a pipeline, and a data-quality screen for Senior+ (both April 2026). |
| Safety research | Simulation and experiment design | Planning a simulation-based safety evaluation (October 2025). |
| Intern | Sometimes | Some 2025 intern loops had a backend design round. Others were trivia plus coding only. |
Recent Zoox system design questions
These are the prompts from 2025 and 2026 loops, newest first. The wording is the rough shape of each question, so expect the interviewer to steer.
- May and August 2026, also 2024: A live robotaxi map with nearby search. Something like ingesting the location of every robotaxi and showing authorized clients a near-real-time map, or finding the moving vehicles near a point that meet some criteria. It came up in a backend technical screen (August 2026) and a software engineer screen (May 2026). A 2024 senior onsite asked for the system behind "their app showing cars on the road."
- August 2026: Ride pricing and payment handling. A backend technical screen. The shape is a pricing service with payment handling that holds up when calls fail, which puts the weight on charging exactly once and reconciling with the payment provider. Matching riders to cars is a different prompt.
- April 2026, data engineers: Teleoperations metrics and fleet data quality. A data engineer onsite asked for a data model to measure how well the teleoperations function works, plus a pipeline to measure vehicle stability across the fleet. A Senior+ screen the same month asked for at least five data-quality checks on a set of tables, with SQL for two.
- October 2025, also 2023: Ride-hailing for the Zoox fleet. A backend onsite in October 2025 had a ride-hailing design. In 2023 an engineering manager was asked to design ride-hailing for the Zoox fleet specifically. The difference from Uber is that the cars belong to Zoox and there's no driver to accept or decline.
ML and safety loops ask their own versions. A safety research screen in October 2025 asked something like choosing which scene parameters to vary in a simulation, how to sample them and what to measure. ML engineers in 2025 had an ML system design round of their own.
How the prompts moved toward the fleet
Older loops asked the generic prompts any company might. None has come back in 2026, so read them as history rather than prep targets:
- A service like TinyURL, asked in a hiring manager round (2022).
- A system for drawing a point or graph chart (2022).
- A document-sharing service for a Staff role in 2023, where the interviewer pasted the requirements and a list of questions and had the candidate answer them one by one.
- Google Scholar, for a sensor software engineer (2024).
- A system design "for a robot" (2019).
- Something like a log aggregation system, in an L3 onsite (2024).
- Something like a heat map, asked by an engineer on the team (2025).
The 2026 prompts have moved on, and the engineer's "won't be like design a Uber/Slack" comment fits that. The format quirk from 2023 is worth remembering, though. If an interviewer hands you a list of questions, answer each one cleanly and in order, then add what the list left out.
How to work a fleet-shaped prompt
Follow-up questions from Zoox design rounds rarely get written down, so the steps below come from the prompts themselves. A live map of moving cars, a nearby search and a payment that must not double-charge all lean on the same few decisions. The live map is the running example.
Pin down who reads it and how fresh it must be
A rider watching their car arrive, an operator watching a whole city and an internal tool replaying yesterday all want different things. Ask which one you're building for. Then settle on numbers: how many vehicles, how often each one reports its position, and how old a dot on the map can get before it's wrong.
Treat locations as a stream that arrives late and out of order
Updates from moving vehicles can arrive late or out of order whenever a connection drops and comes back. Keep the latest position per vehicle by the timestamp the vehicle stamped on it. Arrival order will lie to you. Write raw updates to a log for replay, and serve the map from a small store that holds only the newest position per car.
Index by area for nearby search
Bucket positions by map cell (a geohash or an S2 cell) so "cars near this point" reads a few cells instead of the whole fleet. Cars cross cell lines constantly, so moving a car from one bucket to the next has to be cheap. Say what happens when one cell gets crowded, like an airport pickup zone at rush hour.
Make every money step safe to retry
On the pricing prompt, a client retries, a call to the payment provider times out, and the charge must still happen once. Give each charge an idempotency key, a unique ID that lets the provider spot a repeat. Record what you meant to do before you call out, and run a reconciliation job that compares your records with the provider's.
Say what the user sees when something fails
Name the failure modes out loud: a vehicle goes quiet, a region's server dies, the payment provider is down. For each one, say what the rider or operator sees. A map dot that shows its age is better than one that silently freezes. Put authorization on the map feed too, since the positions of real cars are sensitive.
The worked designs
Four designs are worked end to end in this section: the three fleet prompts from 2025 and 2026, then the data engineer variant. The live robotaxi map is free and the right place to start.
Design a Live Robotaxi Map & Nearby Search
The 2026 prompt with the most reports. Ingest every robotaxi's position, keep a near-real-time map and answer "which cars are near here."
Design Ride Pricing & Payments
A backend screen prompt from August 2026: price a ride and charge for it exactly once, through retries and timeouts.
Design Ride-Hailing for the Zoox Fleet
Matching riders to cars Zoox owns, with no driver in the loop. Backend onsite, 2025.
Design Teleoperations Metrics & Data Quality Checks
The data engineer version from April 2026: a data model, a fleet pipeline and the checks that catch bad rows.
Common questions
Does Zoox have a system design interview?
For most generalist software roles, yes. L3 and L4 loops often have one round, depending on the team, L5 usually has one, and a 2026 Senior/Staff loop had two. Autonomy teams tend to use domain rounds instead, ML engineers get ML system design, and data engineers get a data model and pipeline. Separately, nearly every software screen has an object-oriented coding problem that candidates also call "design."
What system design questions does Zoox ask?
In 2026 the prompts are about the fleet: a live map of robotaxi locations, nearby search over moving vehicles, and ride pricing with reliable payments. Ride-hailing for the Zoox fleet came up in 2025 and 2023. Older loops asked generic prompts like a TinyURL-style service (2022), but nothing like that has been reported recently.
Is the Zoox system design round backend-heavy?
Yes. A frontend engineer who failed it in May 2026 said so, and a frontend-leaning candidate heard the same in July. Expect to go deep on storage, indexing and failure handling whatever your track.
Key Takeaways
- ✓Two rounds get called design: a class-design coding problem in nearly every screen, and a classic system design round in many loops.
- ✓Many L3 to L5 loops have one system design round, and Senior/Staff may have two. Autonomy, ML and data tracks use their own kind.
- ✓The 2026 prompts are about the fleet: a live robotaxi map, nearby search and ride payments. Practice those over generic consumer apps.
- ✓The round is backend-heavy. Go deep on late updates, map indexing, retry-safe payments and what users see when things fail.
Zoox interview guideKnow what Zoox asks before you walk in.
21 coding · 4 system design · 30 behavioral questions
Get their questions first.
InterviewLegend is independent and not affiliated with or endorsed by any company it covers.