Databricks Interview Questions

• Given a list of CIDR blocks, determine whether a given IP address is covered by the ranges — and the inverse: given a start IP and a count of addresses, produce the minimal list of CIDR blocks covering exactly that range (LeetCode 'IP to CIDR' style).Medium
• Implement a variable-sized (N×N) tic-tac-toe board: support placing alternating moves and efficiently reporting whether the last move won, ideally in O(1) per move rather than re-scanning the board (LeetCode 'Design Tic-Tac-Toe' style).Medium
• House Robber — maximize the sum of non-adjacent elements in an array; interviewers often escalate to follow-ups (circular street, or 'rob along a binary tree') as time permits.Easy
• Weighted shortest/cheapest paths in a graph — e.g. find the minimum-cost (and sometimes minimum-time, tie-broken lexicographically) route from source to destination, sometimes with a hop/stop constraint (a graph-optimization variant reported in the phone screen and coding rounds).Medium
• Capacity To Ship Packages Within D Days — find the minimum ship capacity so all packages ship in order within D days (a top-frequency problem on the Databricks LeetCode tag).Medium
• Implement an in-memory snapshotable set / data structure: support add/remove/contains plus snapshot() that freezes the current contents and returns a snapshot id, with iteration over a given snapshot isolated from later mutations.Hard
• Rotting Oranges / Decode String / K Closest Points to Origin / Binary Tree Vertical Order Traversal — representative medium problems pulled directly from the Databricks LeetCode company tag and rotated through the coding rounds.Medium
• Implement a thread-safe bounded blocking queue and use it to build a fixed-size thread pool: producer threads enqueue work, worker threads block waiting for work and process it; the queue must block producers when full and consumers when empty.Hard
• Implement an efficient concurrent logger: many threads submit log messages onto a queue and they must be written out (e.g. in order / without blocking producers), processed by a background consumer.Hard
• Build a thread-safe in-memory key-value store / cache supporting concurrent reads and writes (often framed as 'make this single-threaded structure safe for many threads' or 'design a thread-safe cache').Hard

Reported via Reported on interviewing.io's Databricks guide, the LeetCode Databricks company tag (corroborated via interviewsolver.com, algo.monster, and dsaprep aggregations), and Glassdoor/LeetCode 'Databricks Software Engineer' write-ups. Named problems (IP→CIDR, variable-sized tic-tac-toe, House Robber, weighted paths) come from interviewing.io candidate reports; the LeetCode-tagged list (Capacity to Ship Packages, Rotting Oranges, Decode String, K Closest, Vertical Order Traversal) and the snapshotable-set problem are from the actively maintained Databricks tag and prachub's question bank. · Dedicated concurrency/multithreading round — the most distinctive and, per Blind and interviewing.io, the hardest part of the Databricks loop. The 'efficient logger' problem is named on interviewing.io (corroborated by prachub's 'multithreaded event logger'); the thread-pool / blocking-queue producer-consumer problem is from interviewing.io's write-up; the thread-safe KV store/cache is corroborated by prachub ('Build a Durable Key-Value Cache'). interviewing.io notes these are often NOT on LeetCode, so prep via LeetCode's concurrency section plus real synchronization primitives.