2 coding rounds back to back. 1. LC 1676 LCA of binary tree IV. had to handle the case where nodes dont exist in the tree 2. custom problem: given a stream of SQL queries, build an optimizer that detects redundant subqueries. this was hard af the second one felt more like a real databricks problem. i got a working solution but it wasnt optimal. interviewer hinted at a hash-based approach i didnt fully explore waiting on system design round scheduling

Databricks IC3 coding rounds

efeb1a73

2 coding rounds back to back.

441392e6

the sql optimizer problem is very databricks. they love testing whether you can think about query execution plans. the hash approach is probably building a signature for each subexpression tree and deduplicating

90f2c33b

LC 1676 is medium. if you handled the edge case where nodes dont exist youre fine

6a7d0e4f

databricks system design round will probably be about designing a distributed query engine or a lakehouse architecture. brush up on spark internals and delta lake