Data Engineer interview questions

I've been in data engineering for seven years, the last three in Oman. Started as an ETL developer at an Indian fintech building Informatica pipelines, moved into modern data stack work around 2020 (Spark, Airflow, Snowflake), and for the past three years I've been senior data engineer at an Omani bank building their data lakehouse. Daily I work with Databricks, Apache Spark, Delta Lake, Airflow, and Power BI on the consumption side. I hold Databricks Data Engineer Associate and AWS Big Data Specialty.

Three zones: bronze (raw, immutable, schema-on-read), silver (cleaned, conformed, deduped), gold (business-aggregated for analytics and ML). Each zone has clear ownership and SLAs. Storage in Delta Lake on cloud object storage; processing in Spark via Databricks. Orchestration in Airflow with DAGs versioned in Git. Schema evolution handled explicitly with Delta's schema enforcement; no silent column additions. Data lineage tracked through tools like Unity Catalog or OpenLineage. Cost management: tag everything, monitor query costs in BI tools, archive cold data to cheaper tiers automatically.

First, data classification at ingestion: identify PII columns (national IDs, phone numbers, emails) and mask or encrypt them in non-production environments automatically. For production, the gold layer often needs the PII for legitimate use; we use access controls (Unity Catalog or Ranger) to restrict who can query which columns. Right-to-erasure: design pipelines so a single record can be deleted across all derivatives; this is hard in append-only systems but doable with proper data modelling. Audit logging on PII access. Retention policies enforced through automated cleanup jobs. Privacy is a design constraint, not a post-launch concern.

Pipeline health: Airflow's built-in monitoring for run status and duration, plus a dashboard tracking SLA compliance per DAG. Failed runs page on-call. Data quality: implementation depends on the dataset, but generally I have row-count checks (within tolerance vs prior period), null-rate checks (alert on sudden spikes), business rule checks (revenue can't be negative, dates must be valid), and reconciliation checks (totals match source systems). Tools like Great Expectations or Soda formalise the checks. Critically: I treat data-quality failures as blocking, not warnings. Bad data downstream is more expensive than a delayed report.

I see data engineering as a service function for analysts and scientists, not a gatekeeper. I prioritise their unblocking and try to make their lives easier. I document datasets well so they don't need to ask basic questions repeatedly. I respond fast to feedback when something's wrong. I also push back constructively: if an analyst asks for a one-off SQL query that's actually a recurring need, I'll productise it as a proper dataset in the gold layer instead. The relationship is collaborative; their insights are the value the bank gets from my pipelines.

For a senior data engineer role in Oman banking I'd target OMR 1,700 to 2,100 total package depending on the tech stack and the business context. Banks are willing to pay more because of the regulatory and data-quality requirements. I'm on 60 days' notice. Beyond pay I care about the data maturity of the team; data engineering in an org that doesn't trust data isn't rewarding regardless of pay.