Databricks logo

DevOps Engineer Interview Experience at Databricks2026

Databricks

In Progress 3 rounds Mar 2026 Off Campus
A

Anonymous

3-6 Years · Experienced

General SRE Pattern

No specific, verified real Databricks DevOps candidate account was found. Built from generally reported real DevOps loop patterns. This entry is a reconstruction, not a first-hand transcript.

Application Story

Candidates apply online and are typically contacted for a recruiter screen within a week, followed by a technical round and a system design round, with the hiring manager round scheduled last.

Preparation Journey

Preparation should focus on Kubernetes at scale and cloud infrastructure automation, since Databricks runs large multi-tenant Spark clusters on top of major cloud providers.

Why Selected / Rejected

Pending decision.

Resume Tips

For infrastructure roles at data platform companies, mention any experience running compute-heavy workloads (Spark, ML training jobs). Show knowledge of resource scheduling.

01Round
EASY

Telephonic

30 minutes Remote

Recruiter screen on background and specific interest in infrastructure work at a company whose core product is compute-heavy, bursty Spark and ML workloads rather than a typical steady-state web service.

02Round
MEDIUM

Technical

60 minutes Remote

Covers Kubernetes resource management specifically for compute-heavy, bursty workloads - how to size resource requests versus limits when a job might need 2 CPUs most of the time but spike to 8 during a shuffle-heavy stage, and the risk of setting limits too conservatively (jobs get OOMKilled) versus too generously (poor cluster bin-packing and wasted capacity). The Terraform portion asks candidates to design module structure for provisioning per-team isolated environments with consistent guardrails (mandatory tagging, budget alerts) applied uniformly.

03Round
HARD

System Design

60 minutes Remote

Candidates are asked to design infrastructure for safely running many isolated, resource-intensive workloads on shared Kubernetes clusters without one tenant's job starving another's. A strong approach proposes Kubernetes ResourceQuotas and LimitRanges scoped per namespace as the baseline guardrail, combined with priority classes so that latency-sensitive workloads can preempt lower-priority batch jobs under contention, and node pools segmented by workload profile (memory-optimized versus compute-optimized) rather than a single undifferentiated pool.

Comments

Loading comments...

Questions Asked at Databricks in the DevOps Engineer Interview

Telephonic Round

Recruiter screen on background and specific interest in infrastructure work at a company whose core product is compute-heavy, bursty Spark and ML workloads rather than a typical steady-state web service.

Technical Round

Covers Kubernetes resource management specifically for compute-heavy, bursty workloads - how to size resource requests versus limits when a job might need 2 CPUs most of the time but spike to 8 during a shuffle-heavy stage, and the risk of setting limits too conservatively (jobs

System Design Round

Candidates are asked to design infrastructure for safely running many isolated, resource-intensive workloads on shared Kubernetes clusters without one tenant's job starving another's. A strong approach proposes Kubernetes ResourceQuotas and LimitRanges scoped per namespace as the