Question 1

Discuss a time you had to push back on a requirement.

Accepted Answer

Situation: Stakeholders requested real-time dashboards with sub-second latency on a batch-oriented warehouse. Task: Manage expectations and deliver value without over-investing. Action: I analyzed cost and complexity—real-time would require Kafka, streaming infra, and significant new headcount. I proposed a compromise: micro-batch every 15 minutes with the existing Spark stack, delivering ~90% of value at ~20% of cost. I documented trade-offs (freshness vs....

Question 2

Cloud Composer Overview

Accepted Answer

Why Composer: Managed Airflow on GCP—no cluster ops; integrated with BigQuery, GCS, Pub/Sub. Architectural logic: Creates Airflow environment; DAGs via GCS bucket; automatic upgrades. Supports Airflow 2.x, Celery/K8s executors. When to use: GCP-native pipelines; teams wanting orchestration without ops. Scalability: Auto-scaling workers; environment size configurable. Trade-offs: Cost can add up; less control than self-hosted. Cost: Environment size (CPU, memory) + worker costs....

Question 3

GCP Authentication with Jenkins

Accepted Answer

Architectural options: (1) Service account + JSON key in Jenkins credentials. (2) Workload Identity (Jenkins on GKE)—no keys. Flow: Create SA → download JSON → add to Jenkins → use in GCP plugin or gcloud....

Question 4

Using Service Accounts in GCP

Accepted Answer

**Why Service Accounts (SA)**: Non-human identities for apps, VMs, Cloud Functions. No passwords to rotate; IAM-bound. **Architecture**: Create SA in IAM; assign roles (least privilege). Attach to VM, Cloud Run, Dataflow, Composer. For GKE/Cloud Run: use Workload Identity—pod/container gets SA without key files. Workload Identity Federation extends to AWS/Azure for cross-cloud. **Scalability**: One SA per application/service; avoid shared SAs....

Question 5

About Jira

Accepted Answer

**Jira**: Atlassian project management. Agile: backlogs, sprints, stories. Workflow: To Do → In Progress → Done. **Data engineering use**: Pipeline work, incidents, data catalog updates. Link to code; use labels for discoverability....

Question 6

Discarding Local Changes in Git

Accepted Answer

**Commands**: git restore (unstaged); git restore --staged (staged); git reset --hard HEAD (all). git stash for reversible save. **Why Care**: --hard is destructive; in CI/CD use git clean -fd for untracked. Never --hard on shared branches without alignment. **Pro-Move**: Always git status; git diff before discarding....

Question 7

Explain XComs

Accepted Answer

**What**: XComs (Cross-Communication) in Airflow enable task-to-task data passing. Push via xcom_push() or return; pull via xcom_pull().

**Why**: Pass small metadata—file paths, counts, task IDs. NOT for large payloads—use S3/GCS and pass paths.

**Backend**: DB default; custom (S3, Redis) for scale....

Question 8

Git Bash Commands

Accepted Answer

**Essential**: clone, status, add, commit, push, pull, branch, merge, log, diff, stash, remote. Reset: --soft (undo commit, keep changes); rebase. .gitignore for credentials; Git LFS for binaries. Branch strategies: GitFlow, trunk-based....

Verizon Data Engineer Interview Questions

+5 More Questions with Expert Answers

Other Companies

Verizon Data Engineer Interview Questions

+5 More Questions with Expert Answers

Other Companies