Snap Interview Questions — Page 5

Snap·Software Engineer·Onsite - System Design / Architecture

Apr 2026

System design round at Snapchat for a software engineering role, focused entirely on building an ad-insertion platform for a video streaming service. It was one of the more sprawling design questions I've faced, covering everything from RTB integration to privacy compliance in a single session.

Design an ad-insertion platform for a video streaming service that handles both live and on-demand content.
How would you design the ad decision engine to meet a sub-100ms latency requirement at the edge?
How would you handle fault tolerance and multi-region failover for the ad serving infrastructure?
Walk me through how you'd integrate with real-time bidding exchanges and support ad podding.
How would you implement cross-device deduplication for ad frequency capping?
How would you design the measurement pipeline to track impressions, quartile events, and clicks at scale?
How would you run A/B experiments on ad policies without degrading user experience or revenue?
What privacy compliance considerations would you build into this platform from the start?

“This is the kind of question where you can spend 45 minutes and still feel like you only scratched the surface.”

View Post

Snap·Machine Learning Engineer·Technical Phone Screen

Apr 2026

ML fundamentals interview at Snapchat, went pretty deep on three core topics with a lot of follow-up questions layered on top of each other. The kind of round where you think you're done with a topic and then they just... keep going.

Walk me through the bias-variance trade-off. How do model capacity, training set size, and regularization each affect it, and what does this look like on a learning curve?
How does class imbalance affect model training, and what approaches would you use to address it?
How do you diagnose overfitting in a model, and what's your toolkit for fixing it?

“Started okay on definitions but the follow-ups got granular fast.”

View Post

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

Apr 2026

Snapchat ML engineer interview focused entirely on a deep system design problem around short-video retrieval and recommendation. The question was layered and kept branching into sub-problems, which I wasn't fully prepared for.

Design an end-to-end query-to-video retrieval system for a short-video app, and explain how it fits into a broader recommender stack covering retrieval, ranking, and re-ranking.
How would you represent videos using multi-modal features, and what do you do about the roughly 80% of videos that have no reliable text metadata?
Walk through your choices for offline training, online serving, and ANN indexing for this retrieval system.
How would you evaluate the quality of your retrieval system?
How would you reduce popularity bias in retrieval and recommendation without hurting relevance or engagement?

“I started with a two-tower model and ANN index, which felt right, but I fumbled when they pushed on how retrieval feeds into ranking.”

View Post

Snap·Software Engineer·Onsite - System Design / Architecture

Apr 2026

Snapchat system design round for a software engineer role. The problem was building a Stories feature end to end, which sounds straightforward until you realize how many moving parts they actually want you to cover.

Design a Snapchat Stories system with 24-hour expiry, the ability to save or favorite stories for later re-watching, and offline support where local actions sync back when the user reconnects.
How would you handle media storage and CDN delivery for stories, including expiry and retention policies?
Walk through how you'd model the data for stories, views, and favorites.
How would you design the privacy and access control model for stories?
How do you handle scale for popular creators, specifically the fan-out problem when millions of users need to see a new story?

“This was the whole interview, basically.”

View Post

Snap·Product Manager·Onsite - Product Sense / Strategy

Apr 2026

PM interview at Snap, one question about measuring Snap Map. Short and focused, not a lot of context to go on but the question itself is pretty meaty if you're not prepared for it.

How would you measure the success of Snap Map?

“I started with DAU and engagement time which felt safe but probably too surface-level.”

View Post

Snap·Machine Learning Engineer·Technical Phone Screen

Apr 2026

Snapchat ML engineer screen covering transformer internals. The attention mechanics went fine but positional encoding tripped me up a bit, and the follow-up question exposed the gap pretty clearly.

Can you explain how self-attention works in a transformer, including the Q, K, V structure and how attention scores are computed?
Why does a transformer require positional encoding at all?

“Walked through Q, K, V and the scaled dot-product formula without too much trouble.”

View Post

Snap·Machine Learning Engineer·Technical Phone Screen

Apr 2026

Snapchat ML engineer interview with a tree traversal problem that looks straightforward until you realize you need parent pointers and they don't exist natively in a binary tree. Not a bad round, just required some setup work before the actual logic.

Given a binary tree with unique node values, a target value, and a list of jump offsets, find the target node and then repeatedly walk upward by the number of levels specified in each jump. Return the values of the nodes you land on, stopping early if you'd go past the root.

“First instinct was to just recurse and track depth, but the jumps aren't cumulative from root, they're sequential hops from wherever you currently are.”

View Post

Snap·Machine Learning Engineer·Technical Phone Screen

Apr 2026

ML fundamentals screen for a machine learning engineer role at Snapchat. Five questions covering overfitting, regularization, deep learning, and transformer architecture. Nothing too wild but the transformer questions definitely separated the prep-ers from the wing-it crowd.

What is overfitting, and how would you detect it by looking at training and validation metrics?
What techniques would you use to reduce overfitting specifically in a linear model?
How do you reduce overfitting in a deep neural network? Walk me through your approach.
Explain how self-attention works in a Transformer. What are queries, keys, and values, and how do you actually compute the attention weights?
Why do Transformers need positional information at all, and what are some ways to inject it into the model?

“Pretty standard opener.”

View Post

Snap·Software Engineer·Technical Phone Screen

Apr 2026

Snapchat SWE interview with a geometry/graph problem that looks like a straightforward coordinate check until you realize the real trick is modeling circle overlap as a blocking chain. Not the hardest problem I've seen but the insight required to get there cleanly was non-trivial.

Given a rectangle from the origin to (X, Y) and a list of circles each defined by a center and radius, determine whether a path exists from the bottom-left corner to the top-right corner that stays inside the rectangle, never touches or enters any circle, and only contacts the rectangle boundary at the two corner endpoints.

“I spent the first few minutes trying to think about this as some kind of BFS on a grid which was completely the wrong direction.”

View Post

Snap·Software Engineer·Technical Phone Screen

Apr 2026

Snapchat SWE interview with a combinatorics problem that looks straightforward until you actually have to implement it cleanly. The recursive angle was the main focus.

Given a list of keys and a mapping from each key to a list of values, and a query subset of those keys, return all combinations formed by picking exactly one value per key in order. For example, querying keys [2, 3] from a map like {1: ['a','b'], 2: ['c','d'], 3: ['e','f']} should produce [['c','e'],['c','f'],['d','e'],['d','f']]. Solve it recursively or iteratively using a queue or stack.

“My first instinct was itertools.product and call it a day, but they wanted an actual implementation.”

View Post