Snap interview notes, page 11 of 22

Large Enterprises

Most of these Snap write-ups come from Software Engineers (117), then Machine Learning Engineers (65) and Data Scientists (18). The most reported rounds are the technical phone screen (149), the system design round (45), the behavioral / leadership round (11) and the product sense round (8).

216 interview notes · updated Jul 2026

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

May 2026

Snapchat MLE system design round focused entirely on building a short-video recommendation system. The scope was pretty wide and they pushed hard on the real-time signals angle, which I wasn't fully prepared for.

Design a short-video recommendation system that handles both long-term user preferences and real-time in-session signals like watch time, skips, likes, and follows.
How would you compute and serve real-time session-level features for a user mid-session without blowing up your serving latency?
Walk me through how you'd approach exploration vs exploitation in this recommendation context, especially given freshness requirements for new content.
What online metrics would you track to evaluate whether your recommendation changes are actually working, and how would you structure an A/B test for this system?
How does your design scale to billions of catalog items and hundreds of millions of users?

“This is the whole question, basically.”

View Post

Snap·Machine Learning Engineer·Technical Phone Screen

May 2026

Snapchat ML engineer interview focused entirely on multi-modal retrieval, specifically the text-video/image search space. Five questions, all technical, no behavioral fluff. Felt more like a research discussion than a standard interview loop.

Walk me through how a CLIP-style model works, covering the architecture, how it's trained, and how you'd actually use it at inference time.
What contrastive learning loss functions are commonly used for representation learning? Describe a few and explain when you'd choose one over another.
What are the main drawbacks of using embedding-based retrieval, like a bi-encoder with vector search?
What alternatives exist to pure embedding retrieval, like cross-encoders, hybrid sparse-dense methods, or generative retrieval? What trade-offs do each of them make?
How would you detect and reduce popularity bias in an embedding-based retrieval system?

“This one I felt pretty solid on.”

View Post

Snap·Software Engineer·Technical Phone Screen

May 2026

Snap SWE screen with a pretty basic coding problem. Nothing fancy, just a bit-manipulation type question that you either know or you don't.

Given a list of integers, write a function that returns 1 for each integer that is a power of 2, and 0 otherwise.

“Classic bit trick question.”

View Post

Snap·Product Manager·Onsite - Behavioral / Leadership

May 2026

Snapchat behavioral round, pretty much one meaty question about decision-making under uncertainty. Not a lot to go on from the prompt itself but the question was genuinely hard to answer well.

Describe a situation where you had to make a high-stakes decision without having all the information you needed. How did you figure out what to investigate versus what to just act on, and how did you communicate the uncertainty and trade-offs to others? What happened, and what would you change?

“This one has a lot of moving parts and I think I tried to cover all of them, which was a mistake.”

View Post

Snap·Software Engineer·Technical Phone Screen

May 2026

Snapchat SWE interview with a grid-based pathfinding problem that's trickier than it looks on the surface. The fire spread mechanic adds a layer that makes naive BFS not quite enough.

You're given a 2D grid where cells are either land, wall, or on fire. Fire spreads each minute to adjacent land cells. You start top-left and need to reach bottom-right. What's the maximum number of minutes you can wait at the start before moving and still make it safely? Return -1 if it's impossible, or 10^9 if you can wait indefinitely.

“The edge case about passing through the destination at the same minute fire arrives tripped me up for a bit.”

View Post

Snap·Software Engineer·Technical Phone Screen

May 2026

Snapchat SWE interview with a grid-based coding problem that went deeper than I expected. They weren't satisfied with just a working solution, they pushed hard on complexity, recursion limits, and alternative approaches.

Given an m×n grid of '1's (land) and '0's (water), count the number of islands where cells are connected in all 8 directions. Implement a solution using BFS or DFS, analyze the time and space complexity, and explain how you'd handle very large grids that might hit recursion limits.

“I went with DFS first because it felt more natural to write quickly.”

View Post

Snap·Software Engineer·Technical Phone Screen

May 2026

Snapchat software engineer interview focused on a rate limiter implementation with a bunch of follow-ups around concurrency and distributed systems. The interviewer clearly had opinions about locking strategies and wasn't shy about pushing back.

Implement a rate limiter with an `allow(client_id) -> bool` API that enforces at most N requests per client in any rolling window of duration W.
How would you make the rate limiter thread-safe under concurrent calls, and why is putting a single lock on the shared map a bad idea?
How would you handle memory cleanup for clients that have gone inactive?
How would you extend this to a distributed rate limiter across multiple servers?

“I went with a sliding window using a deque of timestamps per client, which felt natural.”

View Post

Snap·Software Engineer·Technical Phone Screen

May 2026

Snapchat software engineer interview with a bit manipulation problem. Pretty straightforward round, nothing that'll keep you up at night.

Given an integer n, determine whether it is a power of two.

“I went with the bit trick pretty quickly.”

View Post

Snap·Machine Learning Engineer·Onsite - Behavioral / Leadership

May 2026

Behavioral round at Snapchat for an MLE role, focused entirely on innovation. One question, but it had a lot of layers to it and the follow-ups were relentless.

Tell me about a time you introduced an innovative idea, technical or product-level, that meaningfully improved results. Walk through the problem, why existing approaches weren't cutting it, what your insight was, how you validated it, how you got people on board, and what the actual impact was.

“This question sounds like one big prompt but they really do drill into each piece separately.”

View Post

Snap·Software Engineer·Technical Phone Screen

May 2026

Snapchat SWE interview with a grid-based BFS problem that sounds straightforward until you actually try to implement it under pressure. The follow-up angles on binary search tripped me up a bit.

You have a 2D grid where cells are either grass, fire, or walls. Starting from the top-left, you need to reach the bottom-right safehouse. Fire spreads each minute to adjacent grass cells, and walls block it. You can wait at the start before moving. What is the maximum number of minutes you can wait and still safely reach the safehouse? Return 10^9 if you can always make it, -1 if you never can.

“My first instinct was just BFS from start to end, which is wrong because the fire is moving too.”

View Post