← Back to Directory

Databricks interview notes, page 18 of 22

Large Enterprises

Most of these Databricks write-ups come from Software Engineers (192), then Machine Learning Engineers (9) and Data Scientists (8). The most reported rounds are the technical phone screen (136), the system design round (57), the behavioral / leadership round (9) and the onsite coding round (6).

216 interview notes · updated Jul 2026

Databricks·Software Engineer·Technical Phone Screen

Jun 2026
Databricks SWE interview with a data structures design question that had a sneaky concurrency-adjacent twist. Not the hardest problem on paper, but the snapshot semantics tripped me up more than I expected.
  • Design and implement a set data structure that supports creating snapshot iterators. The set should support add, remove, and contains operations, and an iterator() method that returns an iterator over the state of the set at the moment it was called. Subsequent mutations to the set must not affect already-created iterators.

“My first instinct was just to copy the underlying set into a list when iterator() is called, which is correct but I second-guessed myself for a solid two minutes thinking there was some clever persistent data structure angle they wanted.”

View Post

Databricks·Software Engineer·Onsite - System Design / Architecture

Jun 2026
Databricks system design round, one big question that sprawled into like six different topics. They really wanted to see if you could hold the whole thing together without losing the thread.
  • Design a distributed key-value store, covering the API and data model, how you'd partition data, replication and consistency guarantees, failure handling, hot-key problems, persistence, and scaling.

“This one is basically five questions duct-taped together.”

View Post

Databricks·Software Engineer·Technical Phone Screen

Jun 2026
Databricks threw a tree problem at me that looked like a standard traversal question until I read the constraint about not being allowed to define a node class. The whole thing had to be done with raw integer arithmetic on preorder ids, which is a very different beast.
  • Given a k-order Fibonacci tree where nodes are labeled 0 through size-1 in preorder, find the shortest path between two nodes by their ids. You cannot define a node class and must work purely with integer arithmetic based on the tree's structural properties.

“I stared at this for a solid minute because my instinct was to build a tree and do parent pointers.”

View Post

Databricks·Software Engineer·Onsite - System Design / Architecture

Jun 2026
System design round at Databricks for a software engineer role, and it was not a whiteboard-and-move-on kind of session. They wanted pseudocode, not just boxes and arrows, and they grilled me line by line on the message-send and message-deliver paths. Probably the most technically exhausting design interview I've done.
  • Design a chat application end-to-end at the pseudocode level, covering client/server protocol, message ingestion and fan-out, storage and ordering, presence, group vs 1:1 chats, push notifications, and offline delivery.
  • Walk through the pseudocode for the core message-send path and defend it line by line, including correctness and edge cases.
  • Walk through the pseudocode for the message-deliver path and defend edge cases, including scenarios where recipients are offline or on different devices.

“The scope of this thing is brutal if you're not ready for it.”

View Post

Databricks·Data Scientist·Technical Phone Screen

Jun 2026
Databricks data scientist interview with a technical problem that looked straightforward on the surface but had enough edge cases to keep you busy. The core task was a nearest-neighbor style matching problem in SQL or Python, and the real challenge was handling NULLs and deciding whether to standardize features before computing distances.
  • Given two tables of numeric feature vectors, find the top 5 most similar candidate rows for each target row using MSE as the distance metric. Return the target ID, candidate ID, distance, and rank ordered by target then rank.

“The MSE formula itself wasn't the hard part.”

View Post

Databricks·Software Engineer·Technical Phone Screen

Jun 2026
Databricks software engineer interview with a grid-based pathfinding problem that looked deceptively clean on the surface but had enough edge cases to keep you busy for a while. The twist of handling multiple transport modes with separate time and cost constraints made it more interesting than a standard BFS.
  • Given a 2D city grid where each cell is labeled with a transport mode (walk, bike, car, train), blocked cells, a start, and a destination, find the fastest transport mode from start to destination. Each mode can only traverse cells matching its own label. Break ties in minimum time by minimum total cost.

“The problem reads like a normal BFS grid question until you realize each transport mode is essentially its own separate graph traversal.”

View Post

Databricks·Software Engineer·Onsite - System Design / Architecture

Jun 2026
Databricks system design round focused entirely on a concurrent event logger, which sounds scoped but actually opens up into a pretty deep rabbit hole of tradeoffs. Came out feeling okay but not great.
  • Design a thread-safe event logger that supports many concurrent producers. The log() call should be non-blocking or low-blocking, and a background process should flush events to a sink in the order they were submitted or timestamped. Walk through your synchronization choices, how you handle back-pressure when the queue fills, graceful shutdown without dropping events, and how you'd test it under heavy contention.

“I started with the obvious lock-plus-queue approach and the interviewer was fine with it but immediately pushed on what happens when the queue fills.”

View Post

Databricks·Software Engineer·Onsite - System Design / Architecture

Jun 2026
System design round at Databricks for a software engineer role. The whole thing was one big question about building a cloud file sync service, and they really pushed on the details rather than letting you stay high-level.
  • Design a cloud file storage and sync service similar to Google Drive or Dropbox, covering upload and download mechanics, metadata management, versioning, conflict resolution, real-time multi-client sync, sharing and permissions, deduplication, scalability, and consistency.

“This is a lot of surface area for one question.”

View Post

Databricks·Software Engineer·Onsite - System Design / Architecture

Jun 2026
Databricks system design round, one big question about building a brokered bookstore service with no inventory of its own. Pretty meaty problem that touched a lot of ground in one shot.
  • Design a bookstore service that holds no inventory itself. Customers request books, the service fans out to multiple vendors who respond with bids (price, availability, delivery ETA), and then the service ranks and presents the best offers. Cover vendor onboarding and catalog normalization, request fan-out with timeouts and partial results, bid scoring, order placement and settlement, and reliability under slow or failing vendors at high QPS.

“This one sprawled in every direction and I underestimated how much ground there was to cover.”

View Post

Databricks·Data Scientist·Technical Phone Screen

Jun 2026
Databricks data scientist interview with a statistics-heavy question on hypothesis testing. Pretty straightforward if you've done any stats work, but the level of rigor they expected was higher than I anticipated.
  • You flip a coin 1000 times and observe 560 tails. At a 0.05 significance level, test whether the coin is fair. Walk through your hypotheses, test statistic, p-value calculation, and conclusion.

“I knew this was a two-sided z-test for a proportion but I fumbled the variance formula for a second.”

View Post