← Back to Directory

Nuro interview notes, page 2 of 4

Mid-sized

Most of these Nuro write-ups come from Software Engineers (23), then Machine Learning Engineers (9) and Backend Engineers (2). The most reported rounds are the technical phone screen (17), the system design round (13), the behavioral / leadership round (2) and the hiring manager screen (1).

35 interview notes · updated Sep 2026

Nuro·Software Engineer·Onsite - Behavioral / Leadership

May 2026
Did a behavioral round at Nuro for a software engineer role, pretty much just a standard resume walkthrough and project deep dive. Nothing too wild.
  • Walk me through a project on your resume in depth.

“Pretty standard stuff.”

View Post

Nuro·Software Engineer·Onsite - System Design / Architecture

May 2026
Nuro system design round, pretty deep into autonomous vehicle data infrastructure. One big question that sprawled across like six different sub-topics. Left feeling okay about it but definitely glossed over some parts I wish I hadn't.
  • Design a system that ingests petabyte-scale raw logs from autonomous vehicles, supports multi-modal data (sensor feeds, perception outputs, video), and lets engineers search for specific scenarios like 'find all clips where the AV saw a pedestrian near a stop sign'. Cover ingestion, feature extraction at ingest time, indexing strategies, query API design, storage tiering, and access control.

“This was basically the entire interview wrapped into one prompt.”

View Post

Nuro·Software Engineer·Technical Phone Screen

May 2026
Nuro coding screen, one question the whole time. It's a merge intervals variant but flipped: find the gaps instead of the merged chunks. Felt straightforward once I saw it but I definitely over-thought the setup before realizing it was just a sweep.
  • Given a list of intervals, return all the gaps between them: the maximal ranges on the number line not covered by any input interval.

“I recognized the merge intervals pattern pretty fast, which actually slowed me down a bit because I kept second-guessing whether they wanted something fancier.”

View Post

Nuro·Software Engineer·Technical Phone Screen

May 2026
Nuro SWE interview had a Minesweeper reveal problem that looked straightforward until I actually had to think through all the edge cases in the BFS expansion. Not a bad problem, just one where the details bite you if you're moving fast.
  • Implement the reveal logic for a Minesweeper board. Given a grid where cells can be unrevealed mines, unrevealed empty cells, or various revealed states, and a click coordinate, return the updated board following the standard rules: clicking a mine reveals it as exploded, clicking an empty cell with no adjacent mines triggers a flood-fill expansion in all 8 directions, and clicking an empty cell with adjacent mines just shows the count.

“The BFS expansion part is where I slowed down.”

View Post

Nuro·Machine Learning Engineer·Onsite - System Design / Architecture

May 2026
System design round at Nuro for an ML Engineer role, focused entirely on designing a production inference pipeline end to end. Pretty intense scope for a single session, covering everything from request routing to canary deploys to drift monitoring.
  • Design a production ML inference pipeline that serves predictions to downstream applications, covering request routing, model storage and deployment, feature computation, latency and availability, monitoring, and safe rollout strategies.
  • How would you handle rollback, A/B testing, and canary deployments when releasing a new model version?
  • How do you monitor model quality, data drift, latency, error rates, and resource usage in a live inference system?

“This was basically the whole interview in one prompt.”

View Post

Nuro·Software Engineer·Onsite - System Design / Architecture

May 2026
Nuro system design round for a software engineer role, focused entirely on building a job scheduler from scratch. The discussion went deep pretty fast, covering concurrency, fault tolerance, and architectural tradeoffs. Left feeling like I could've structured the scaling conversation better.
  • Design and implement a scheduler that triggers jobs based on a configured frequency, such as every N seconds or milliseconds, with multiple jobs running concurrently.
  • How do you handle thread safety when multiple jobs are running concurrently in this scheduler?
  • What happens when a job misses its scheduled tick, for example if the system is overloaded or the job itself runs long?
  • How would you implement graceful shutdown for this scheduler so in-flight jobs complete without accepting new work?
  • Compare a thread-per-job design against a single-threaded priority queue or timer wheel approach. What are the real tradeoffs?

“Started with a basic thread-per-job model because it felt intuitive, but the interviewer kept pushing on what happens when you have thousands of jobs.”

View Post

Nuro·Machine Learning Engineer·Technical Phone Screen

May 2026
Interviewed for an ML Engineer role at Nuro and got a tree problem that looked straightforward until I actually tried to code it up. The N-ary twist on what's basically a path sum question kept me second-guessing my recursion logic the whole time.
  • Given the root of an N-ary tree where each node has an integer value and any number of children, find the maximum sum of values along any path in the tree. The path can start and end at any two nodes and cannot revisit nodes.

“My first instinct was to treat it like the binary tree diameter problem, where you track the best single-arm extension you can pass upward and combine two arms at each node.”

View Post

Nuro·Software Engineer·Technical Phone Screen

May 2026
Interviewed for an infrastructure role at NURO and it was pretty much a deep dive into Kubernetes and Docker internals the whole way through. They clearly wanted someone who actually understood what was happening under the hood, not just someone who could recite kubectl commands.
  • Can you explain the difference between a Pod, a Deployment, and a Service in Kubernetes, and when you'd use each?
  • How does the Kubernetes scheduler decide which node to place a pod on?
  • Walk me through how networking works in Kubernetes, covering CNI, Services, and Ingress.
  • How do Persistent Volumes and Persistent Volume Claims work, and what's the difference between static and dynamic provisioning?
  • How does Docker image layering work, and what actually happens during a build?
  • What is the relationship between containerd and runc, and how do they fit into the container runtime stack?
  • How do Linux namespaces and cgroups work together to isolate containers?
  • How does a rolling update work in Kubernetes, and what controls the pace of it?
  • What are the different ways to scale workloads in Kubernetes, and what are the tradeoffs between them?
  • A pod is stuck in CrashLoopBackOff. Walk me through how you'd debug it.

“Started fine but I rambled too long on pods and didn't leave enough space to talk about how Services abstract the networking.”

View Post

Nuro·Machine Learning Engineer·Onsite - System Design / Architecture

May 2026
Interviewed at Nuro for an ML Engineer role and got blindsided on the first round. HR said it was ML design, so I prepped the wrong thing entirely and walked in completely unprepared for what they actually wanted.
  • Design an end-to-end inference pipeline for a machine learning system.

“Prepped for ML design stuff like model selection, training strategies, feature engineering.”

View Post

Nuro·Software Engineer·Technical Phone Screen

May 2026
Nuro data engineer interview with a pretty meaty technical problem centered on autonomous driving clip data. The core task was merging overlapping time intervals and computing deduplicated coverage, and you had to do it in both SQL and Python. Felt like a real problem they actually deal with, not some contrived leetcode thing.
  • Given a table of driving clip intervals per run (with overlaps, duplicates, and gaps), write a SQL query to compute the total deduplicated time covered per run_id after merging all overlapping intervals.
  • Write a Python solution that takes a list or dataframe of intervals per run_id and computes the same deduplicated cumulative time as the SQL query.

“This is the kind of problem that looks approachable until you realize SQL doesn't have a native 'merge intervals' primitive.”

View Post