Netflix Interview Questions — Page 12

Netflix·Software Engineer·Technical Phone Screen

May 2026

Netflix coding screen for a software engineer role, basically one problem with three increasingly annoying rule levels bolted on. The kind of question that feels simple until you're twenty minutes in and realizing you missed an edge case from level one that now breaks level three.

Implement a string-to-integer parser from scratch, without using any built-in conversion methods, across three rule levels: first handling optional leading minus signs and throwing on overflow or non-digit characters; then tightening validation for any special character; then adding strict leading-zero rules where only a single zero prefix is allowed.

“The first level felt manageable.”

View Post

Netflix·Data Scientist·Technical Phone Screen

May 2026

Netflix data scientist screen, heavy on stats and experiment design fundamentals. Six questions back to back, no coding, just pure probability and methodology. Felt more like a grad school oral exam than a job interview.

State Bayes' rule and apply it: given a disease prevalence of 1%, a test sensitivity of 99%, and a false positive rate of 5%, what is the probability of actually having the disease given a positive test result?
In an observational study, why does adding control variables sometimes change the coefficient on the main variable you care about? And when can adding controls actually make things worse?
State the Central Limit Theorem and explain what it means practically for the sampling distribution of the sample mean.
For X drawn from a uniform distribution on the interval (a, b), derive the mean and variance.
Write out the one-sample t-statistic for testing whether the population mean equals some hypothesized value, given a sample mean, sample standard deviation, and sample size.
Define effect size and Minimum Detectable Effect. How do statistical power, variance, sample size, and significance threshold each affect the MDE?

“I knew the formula but the number still surprised me when I computed it.”

View Post

Netflix·Software Engineer·Onsite - System Design / Architecture

May 2026

Netflix ads platform system design, one big open-ended question that expanded in every direction I tried to close it off. Walked out unsure if I'd gone too deep on the data model and not deep enough on the operational stuff.

Design the full data model for an ads platform, covering core entities like advertisers, campaigns, ad groups, creatives, targeting, budgets, and bidding, plus the fact and dimension tables needed for analytics, and how the model supports billing, attribution, reporting, pacing, and frequency capping.

“This question just kept growing.”

View Post

Netflix·Software Engineer·Technical Phone Screen

May 2026

Netflix SWE interview that went deep on a sliding window variant I thought I knew cold. The follow-up about dynamic insertions turned a familiar problem into something I hadn't really thought through before.

You have a sequence and a longest-unique-subarray solution. Now support inserting an element at an arbitrary position in the middle of the sequence. How does this affect your sliding window state, and what data structures let you handle this efficiently?
Walk through the time and space complexity after adding dynamic insertion support to the longest-unique-window problem.
Write and run test cases covering: insertion at the head, insertion at the tail, insertion inside an existing unique window, and insertion across a duplicate boundary.

“I started okay, talking through how a standard sliding window breaks down the moment you touch the middle of the sequence.”

View Post

Netflix·Software Engineer·Onsite - System Design / Architecture

May 2026

Netflix system design round, one big question about an ad platform data model. Took up the whole session and went pretty deep into schema design, indexing, and the different flows. Felt okay about it but there were definitely gaps I noticed only after.

Design a logical data model for an advertising platform, covering advertisers, campaigns, creatives, targeting, user identifiers, and delivery logs. Explain key entity relationships, how the schema supports the creation, serving, and tracking flows, and what indexing or partitioning decisions you'd make at scale.

“This is a bigger question than it looks.”

View Post

Netflix·Software Engineer·Onsite - System Design / Architecture

Jun 2026

Netflix system design round focused entirely on ads platform data modeling. Pretty deep dive, they wanted real schema decisions not just hand-waving at entity relationships.

Design the data model and schema for an ads platform, covering advertisers, campaigns, ad groups, creatives, impressions, clicks, conversions, and budget tracking.
How would you design the schema to support analytical queries like CTR, spend, and conversion rate broken down by various dimensions?
Compare star schema versus snowflake schema for this ads platform use case. Which would you choose and why?
How would you separate OLAP and OLTP concerns in this system? What goes where and why?

“I started with the advertiser and campaign hierarchy which felt natural, but then I got kind of lost deciding how granular to make the budget entity.”

View Post

Netflix·Software Engineer·Onsite - System Design / Architecture

Jun 2026

Netflix system design round for a software engineering role. The whole thing was centered on one big problem: design an ad pacing system for a large-scale streaming platform. Dense, multi-layered, and honestly the kind of question where you realize halfway through how much surface area there actually is.

Design an advertising pacing system for a large-scale video streaming platform. Campaigns have budgets, flight dates, and delivery goals. The system receives millions of ad requests per second and must decide which campaigns are eligible to serve while spreading spend smoothly over the campaign lifetime rather than burning budget early.
How do you prevent overspending when hundreds of serving nodes are all making spend decisions for the same campaign at the same time?
How do you handle nested budget scopes, where a single ad impression can count against an ad-group budget, a daily budget, and a lifetime campaign budget all at once?
The traffic forecast turns out to be badly wrong due to an unexpected spike. How does the pacing controller react, and what prevents over-delivery before the next control loop tick?
An advertiser cuts a campaign's daily budget in half mid-flight. How quickly does that change propagate to every serving node, and how do you prevent overspend during the propagation window?
One campaign accounts for a huge share of total spend and is concentrated in a single hot region. How do you pace it without that campaign's counter becoming a bottleneck for the whole serving fleet?
How would you layer frequency capping (per-user impression limits) on top of the pacing system, and how do the two interact when making eligibility decisions?

“This one took me a few minutes to even scope properly.”

View Post

Netflix·Software Engineer·Technical Phone Screen

Jun 2026

Phone screen for a full-stack role at Netflix, just one round with a topological sort problem. Passed it.

Solve a topological sort problem.

“Pretty classic graph question for a phone screen.”

View Post

Netflix·Software Engineer·Technical Phone Screen

Jun 2026

Netflix coding round, one meaty question about implementing a JSON path query function. Felt like it started simple and then kept growing as they pushed on edge cases and design tradeoffs.

Implement a JSON path query function that takes a nested JSON object and a dot-separated path string, returning all matching values. Path segments can be literal keys, array indices, or '*' wildcards that match any single key or index at that level. How do you handle consecutive wildcards, missing keys, non-object intermediates, and empty results? Also discuss recursion vs iterative DFS and when to short-circuit.

“The wildcard part is where I started to feel the pressure.”

View Post

Netflix·Software Engineer·Onsite - System Design / Architecture

Jun 2026

Netflix ops-style system design round where they throw a target system at you and want you to tear it apart live. Less about drawing boxes and more about knowing what breaks and why.

Given a Kafka cluster, where are the likely bottlenecks, what happens when a broker node fails, and how does failover work?
Walk through a Spark batch job end to end. Where can it bottleneck and what failure modes should you plan for?
For a distributed job scheduler, how would you handle node failures and ensure jobs don't get dropped or double-executed?
How do you balance consistency, availability, and scalability across these systems? Walk through the tradeoffs for each.
Explain how core building blocks like consistent hashing, sharding, replication, caching, and queueing apply to the systems you've discussed.

“I started with partition lag and consumer group rebalancing, which felt right, but then fumbled a bit on exactly how leader election works when a broker drops.”

View Post