Home/Interview prep/Data Engineering: Spark, Databricks & Microsoft Fabric

Walk into the interview already prepared.

Timed mock tests and model answers written by working engineers, for every level from fresher to lead. Pick any of our 75 courses below.

Topic: Data Engineering (Spark, Databricks, Fabric, Airflow). Questions are shared by related courses. View the Data Engineering: Spark, Databricks & Microsoft Fabric course → · Follow the Data Engineer roadmap and salary guide →

All Data Engineering: Spark, Databricks & Microsoft Fabric interview questions at a glance+

Fresher (0–3 years): Junior Data Engineer

  1. Design a simple daily pipeline from an API to a dashboard.
  2. Explain batch vs streaming processing.
  3. How do you check data quality?
  4. ETL vs ELT: in ELT data is…
  5. Apache Spark is mainly used for…
  6. Which file format is columnar and efficient for analytics?
  7. Apache Airflow is used to…
  8. An idempotent pipeline means…
  9. A data lakehouse combines…

Mid-level (3–8 years): Data Engineer

  1. How do you optimise a slow Spark job?
  2. Design a medallion (bronze/silver/gold) architecture.
  3. A daily pipeline failed overnight and dashboards are stale. What do you do?
  4. A Spark job is slow because one task runs much longer. Likely cause?
  5. Broadcast join is useful when…
  6. Delta Lake adds which capability to data lakes?
  7. Change data capture (CDC) is…
  8. Data lineage helps…
  9. Partitioning a large table by date mainly…

Experienced (8–15 years): Senior Data Engineer / Data Architect

  1. Design a data platform for a company moving from spreadsheets to analytics.
  2. How do you implement data governance without slowing teams down?
  3. How do you migrate an on-prem Hadoop/ETL estate to the cloud?
  4. Data mesh mainly means…
  5. Data contracts between producers and consumers…
  6. Lakehouse cost control at scale mainly comes from…
  7. Choosing between Databricks, Fabric and Snowflake should weigh…
  8. GDPR/DPDP "right to erasure" in a lakehouse requires…
  9. Exactly-once streaming results usually need…