Snowflake Interview Questions — Page 3

Snowflake·Software Engineer·Technical Phone Screen

Apr 2026

Snowflake SWE interview with a tic-tac-toe board validation problem. Trickier than it looks on the surface because of all the edge cases around turn counts and the 'stop after win' rule. Decent problem overall, felt like it was testing whether you actually think through state machines rather than just coding the obvious.

Given a 3x3 tic-tac-toe board represented as an array of 3 strings, write a function that returns the game status: whether the board is invalid, X wins, O wins, a draw, X's turn, or O's turn.

“I started with the easy stuff, count X's and O's, check that X has either the same count as O or one more.”

View Post

Snowflake·Software Engineer·Onsite - System Design / Architecture

Apr 2026

Snowflake system design round, one question about building an ACL service. Pretty focused interview, no fluff, just got handed a fairly open-ended design problem and had to run with it.

Design a scalable ACL (Access Control List) service that handles authorization checks on behalf of another service. It should determine whether a given user or client is permitted to perform specific actions on specific resources.

“I started with the data model since that felt like the core of it: who, what action, what resource.”

View Post

Snowflake·Software Engineer·Onsite - System Design / Architecture

Apr 2026

Did a system design round at Snowflake for a software engineer role. Came in expecting to build something from scratch but the question was more about diagnosing an existing setup than designing one, which threw me off a bit.

Given a basic system that stores user passwords, what are the potential bottlenecks?

“Walked in thinking this was a full design-from-scratch exercise, so I started rambling about schemas and storage layers before realizing they just wanted me to poke holes in something already built.”

View Post

Snowflake·Software Engineer·Technical Phone Screen

Apr 2026

Snowflake software engineer phone screen, one hour, pretty much the entire session was a single hard DP problem and its followup. Brain fog hit at the worst possible time and I couldn't close it out.

Given a list of events each with a start day, end day, and value, and a maximum number of events you can attend, find the maximum total value you can collect. You can only attend one event at a time, must attend it in full, and two events cannot share a boundary day.

“My brain just was not there.”

View Post

Snowflake·Software Engineer·Technical Phone Screen

Apr 2026

Snowflake coding round with a file system design problem. Pretty classic but there are enough edge cases to trip you up if you're not careful about your data structure choices.

Design a simplified in-memory file system that supports listing files and directories, creating directories recursively, adding content to files, and reading file content.

“My first instinct was to use a trie-like structure and I think that was the right call, but I wasted a few minutes second-guessing whether to just use nested hashmaps.”

View Post

Snowflake·Software Engineer·Technical Phone Screen

Apr 2026

Snowflake SWE interview with a pretty meaty coding question around virtual functions and building out a SQL engine. Not a typical leetcode grind, more of a design-meets-implementation problem that caught me a bit flat-footed.

Using virtual functions, design and implement a database SQL engine that supports a set of query operators.

“This one had more depth than I expected.”

View Post

Snowflake·Technical Program Manager·Onsite - Behavioral / Leadership

Apr 2026

One question from what looks like a Snowflake TPM interview, behavioral flavor, asking about your impact on a program. Not much else to go on here.

In what ways has this program gotten better or worse because of your direct involvement?

“This is the kind of question that sounds easy until you actually try to answer it without sounding like you're just listing accomplishments.”

View Post

Snowflake·Software Engineer·Technical Phone Screen

Apr 2026

Snowflake coding interview with a tree pruning problem that looked straightforward until the follow-up on complexity hit. Not the worst session I've had but definitely left me second-guessing my DP formulation.

Given a rooted tree and an integer N, find the minimum number of nodes to delete so that the tree's maximum depth does not exceed N. When a node is deleted, its entire subtree is removed. Return the minimum number of deletions and discuss the complexity of your approach.

“I went with DFS to compute subtree depths first, which felt right, but then I stumbled when trying to articulate the greedy vs DP angle.”

View Post

Snowflake·Software Engineer·Technical Phone Screen

Apr 2026

Snowflake software engineer interview that went deep on graph algorithms and scheduling. Three connected parts to one problem, and by the third part I was genuinely winging it.

You have N services with dependency relationships (u must start before v). Given these pairs, produce a valid startup order or detect and report a cycle if one exists.
Using the same dependency graph, group the services into startup layers where all services with no unmet dependencies form the first layer, then the next set, and so on.
Each service has a startup time and you have K CPU cores available. Schedule the startups to minimize total completion time while respecting the dependency constraints. Describe your algorithm, its complexity, and any practical heuristics for parallelism.

“Topological sort, so I knew the shape of it pretty fast.”

View Post

Snowflake·Software Engineer·Onsite - System Design / Architecture

Apr 2026

Snowflake system design round for a software engineer role. The whole session was basically one big distributed systems question with a bunch of follow-ups that kept drilling deeper. Not a comfortable interview if you haven't thought carefully about scheduler correctness before.

Design a distributed cron job scheduler that triggers user-defined jobs on recurring schedules. Worker capacity is unlimited, so the focus should be on correct scheduling, state management, and reliable triggering rather than scaling the compute side.
What clarifying questions would you ask before starting the design? Things like schedule granularity, timezone support, backfill behavior on resume, at-least-once vs at-most-once delivery, and what the job hand-off target looks like.
Walk through the data model that drives scheduling correctness. What fields does a job record need, and why?
Describe the core scheduler loop: how does it find due jobs efficiently at a scale of a million job definitions, claim them safely, emit a run record, and advance the schedule?
With multiple scheduler replicas all seeing the same due jobs simultaneously, how do you prevent the same job from firing more than once per tick?
What are the exact semantics of pause and resume, and how do you handle the race where a scheduler has already picked up a job for the current tick at the exact moment pause commits?
If a scheduler instance crashes after writing the run record to the database but before the run reaches the queue, how does your design ensure the trigger is neither lost nor delivered twice?
A job is configured to run every minute but each execution takes about five minutes. How do you handle the overlap: skip the new run, queue it, or allow concurrent runs? How does the data model express that policy?
The metadata store becomes a throughput bottleneck with a million jobs all being scanned and updated. How do you partition or shard the scheduling work across replicas without losing the no-double-fire guarantee?
How would you handle per-job timezones and make sure a daylight saving transition doesn't cause a job scheduled for 2 a.m. to either fire twice or get skipped entirely?

“This is the kind of question where the unlimited workers constraint is a trap if you don't internalize it fast.”

View Post