Google Data Engineer Mock Interview

Tell me about a time you had to redesign a data pipeline at Google-scale (e.g., improving latency/cost/reliability). What was the original design, what signals did you use to diagnose the problem, and how did you validate the new approach before rolling it out to production?

You’re given event data from Google Analytics/Ads-style streams and asked to build a near-real-time pipeline that supports both ad hoc exploration and consistent reporting. How would you design the schema, partitioning strategy, and data quality checks (including handling late/out-of-order events) using tools such as BigQuery/Dataflow/Dataproc?

Imagine you discover that a critical metric used by product teams is wrong due to a subtle data correctness issue (e.g., join duplication or changed business logic). Walk me through how you would debug the issue, identify the root cause, communicate impact, and implement a durable fix with tests/monitoring.

A stakeholder wants a new dashboard “today,” but the underlying data sources are inconsistent across regions and there’s no clear single source of truth. How would you align requirements, define canonical entities/metrics, and ensure the pipeline produces trustworthy results for all regions?

In a system where multiple teams publish to your shared warehouse, you’re seeing escalating costs and slow query performance. Describe how you would investigate the bottleneck (partitioning/clustering, file sizes, caching/materializations, query patterns), propose measurable improvements, and manage rollout to avoid breaking downstream users.