← Back to Directory

Snowflake interview notes, page 12 of 14

Large Enterprises

Most of these Snowflake write-ups come from Software Engineers (118), then Data Scientists (12) and Backend Engineers (2). The most reported rounds are the technical phone screen (91), the system design round (30), the behavioral / leadership round (5) and the online assessment (2).

135 interview notes · updated Oct 2026

Snowflake·Software Engineer·Onsite - Multi Round

Jun 2026
Snowflake software engineer interview with a deep technical and organizational deep-dive on your most challenging project. The question has a lot of sub-parts and they will push hard on specifics, so vague answers won't survive long.
  • Walk me through your most challenging project end to end, covering the problem, your specific role, the key decisions you made, technical and organizational obstacles, cross-functional work, a conflict you navigated, how you measured success, and what you'd do differently.

“This is basically nine questions wearing a trench coat.”

View Post

Snowflake·Software Engineer·Technical Phone Screen

Jun 2026
Snowflake technical phone screen with a tree DP problem that looked like a standard tree traversal until you realized it was basically a knapsack on a tree. Harder than I expected for a phone screen.
  • Given a rooted tree where each node has a non-negative weight, you can delete any node along with its entire subtree (but not the root). With a total weight budget W for kept nodes, maximize the depth of the resulting tree. Return the max depth and which nodes to delete.

“I spent the first few minutes thinking about greedy approaches and they all fell apart pretty fast.”

View Post

Snowflake·Software Engineer·Technical Phone Screen

Jun 2026
Snowflake SWE interview with a graph/topological sort problem that had a twist I wasn't fully prepared for. The core question was straightforward but the follow-up about grouping tasks by levels added enough complexity to make me sweat.
  • Given N tasks labeled 0 to N-1 and a list of prerequisite pairs, return a valid execution order of all tasks (or an empty list if none exists). Also group the tasks by levels, where each level contains tasks that can only begin once all tasks in earlier levels have finished. Walk through your algorithm, its complexity, and how you handle cycle detection.

“I knew Kahn's algorithm well enough to get the basic ordering out, but the level-grouping part tripped me up for a minute.”

View Post

Snowflake·Software Engineer·Technical Phone Screen

Jun 2026
Snowflake SWE interview with a path filtering problem that sounds deceptively manageable until you're actually in it trying to talk through edge cases and complexity tradeoffs at the same time.
  • Given a list of slash-delimited hierarchical path strings and a list of paths to delete, remove every path that matches or falls under any delete path. Return the surviving paths. Design an efficient solution, handle edge cases like duplicates and trailing slashes, and analyze time and space complexity.

“My first instinct was to sort both arrays and do prefix-skipping, which works, but I fumbled explaining why that's actually efficient.”

View Post

Snowflake·Software Engineer·Technical Phone Screen

Jun 2026
Snowflake software engineer interview with two back-to-back technical problems: a hierarchical in-memory file system and a fixed-size circular queue. Both were deeper than they looked on the surface, especially the file system part which had a lot of edge case discussion baked in.
  • Design and implement an in-memory hierarchical file system supporting ls, mkdir, createFile, appendToFile, and readFile operations on absolute UNIX-style paths, with strict error handling (no auto-creation of missing intermediate directories). Also analyze time/space complexity and discuss edge cases around root and invalid paths.
  • Implement a fixed-size circular queue with O(1) time for all operations (enQueue, deQueue, Front, Rear, isEmpty, isFull), and justify your data structure choice versus alternatives like a doubly linked list.

“This one ate up most of the session.”

View Post

Snowflake·Software Engineer·Technical Phone Screen

Jun 2026
Snowflake software engineer interview with a meaty tree problem that had more moving parts than I expected. The question looked like a straightforward traversal until they started asking about overflow handling and serialization helpers on top of it.
  • Given two complete binary trees with identical structure, replace every node value in the second tree with the sum of all values in the corresponding subtree of the first tree. Implement this in O(n) time and O(h) extra space, and also write helpers to build a tree from a level-order array and serialize it back to an array for verification.

“I went straight for a recursive post-order traversal and got the core transformation working pretty quickly.”

View Post

Snowflake·Data Scientist·Technical Phone Screen

Jun 2026
Snowflake DS interview that was basically a full experiment design case study. One long question broken into five parts covering everything from sample size math to ramp decisions. Dense but fair if you've actually run A/B tests before.
  • You're replacing a feed ranker with a new model. Baseline CTR is 2%, you expect a +5% relative lift, and you want 90% power at alpha 0.05. Daily traffic is 1M users on a 50/50 user-level split. Compute the required sample size and minimum test duration, adjusting for 5% bot traffic and 1% sample ratio mismatch.
  • What guardrail metrics would you set for this experiment, and how would you use sequential monitoring with alpha-spending to allow early stopping without inflating Type I error?
  • How would you use CUPED to address novelty effects and day-of-week variation? What exact covariate would you use, and how do you validate that variance reduction isn't introducing bias?
  • How do you prevent interference and contamination between experiment arms, and when would you consider a geo-based or holdout design instead of user-level bucketing?
  • After the experiment concludes, how do you verify that you actually achieved the intended power, check for heterogeneous treatment effects across user cohorts, and decide when to ramp to 100%?

“This is where I spent most of my mental energy.”

View Post

Snowflake·Software Engineer·Onsite - System Design / Architecture

Jun 2026
Snowflake system design round, one big question that sprawled in every direction. The scope was way larger than I expected and I spent too long on the easy parts before realizing I hadn't touched half the requirements.
  • Design a caching strategy for a materialized view system where views are organized as a DAG with dependencies on each other. Cover what to cache, where to cache it, cache key design and versioning, eviction, freshness guarantees, and how to handle invalidation when base data changes. Also explain how updates propagate through the DAG, how you'd deal with partial failures and backfills, hot key problems, and how you'd monitor correctness and performance.

“I started with the easy stuff, in-memory vs disk vs distributed, and burned maybe 10 minutes on that before the interviewer nudged me toward the DAG propagation piece, which was clearly the heart of it.”

View Post

Snowflake·Software Engineer·Onsite - System Design / Architecture

Jun 2026
System design round at Snowflake for a software engineering role. The whole session was one big open-ended problem about building an internal SDK to abstract away REST calls across a large org. No performance numbers given, which was both freeing and a little disorienting.
  • Design an internal SDK that lets application developers call internal REST services through typed methods instead of writing HTTP boilerplate by hand. The layer should handle retries, auth, observability, and versioning transparently.
  • How would you expose the typed caller-facing API, and what's the tradeoff between generating clients from a contract ahead of time versus resolving calls dynamically at runtime?
  • What controls should callers still have even if the SDK hides HTTP verbs and status codes? What breaks if every failure maps to a single generic exception?
  • How would you handle retries safely when some endpoints are mutating writes rather than idempotent reads?
  • How do you handle service-to-service authentication without making callers manage secrets themselves?
  • A downstream team ships a breaking API change without bumping the version and breaks callers in production. What in your design should have caught this, and how do you recover?
  • How would you extend the SDK to support streaming or long-running async calls without breaking the existing typed request/response model?
  • If you moved all the cross-cutting logic into a service mesh sidecar instead of an in-process library, what do you gain and lose? Which concerns can only live in the in-process SDK?

“This is the kind of question where you can spend 40 minutes and still feel like you only scratched the surface.”

View Post

Snowflake·Software Engineer·Technical Phone Screen

Jun 2026
Snowflake SWE interview with a pretty gnarly tree DP problem. The core concept isn't too bad but the deletion semantics made the recurrence way harder to think through cleanly under pressure.
  • You have an N-ary tree where deleting an interior node promotes its children up to the deleted node's parent (order preserved), and the root can never be deleted. Given the root and an integer K representing the maximum allowed height in edges, find the minimum number of node deletions to bring the tree's height down to at most K.

“The deletion-promotes-children part is what gets you.”

View Post