Technical screening questions
Ask for one example per question and listen for specifics.
- What is the medallion architecture? A good answer: Bronze, silver and gold layers and why data is refined in stages.
- How do you handle small files and skew? A good answer: Optimize, partitioning and salting.
- How do you control cost on Databricks? A good answer: Cluster sizing, auto-termination and job clusters.
- How do you handle schema changes? A good answer: Schema evolution and expectations.
- How do you secure data? A good answer: Unity Catalog permissions and masking.
Quick screening questions
Use these when you have only a few minutes.
- How do you structure a Delta Lake pipeline from raw to curated?
- How have you tuned a slow Spark job on Databricks?
- How do you manage access with Unity Catalog or equivalent?
What separates the levels
Match the level to the requirement before you submit.
- Junior: writes transformations.
- Mid: owns pipelines and tuning.
- Senior: designs the lakehouse and governance.
Red flags
Any of these is a reason to probe further.
- Only notebooks, no scheduled jobs.
- Cannot discuss cluster cost.
- No Delta Lake experience.
Tools you should hear mentioned
A Databricks engineer with real experience usually names several of these without prompting.
- Databricks
- PySpark
- Delta Lake
- Unity Catalog
- Workflows
- dbt or SQL
Run the call well
Ask for the last project first, then dig into the second or third question. Take notes on versions, team size and outcomes, and record them with the submission so the vendor’s later technical interview does not surprise you.
Frequently asked questions
How long should a Databricks engineer screening call take?
Fifteen to twenty minutes is enough to cover the project, three technical questions and availability, rate and work authorization.
Do recruiters need to know the technology?
Working vocabulary is enough. Use the sourcing guide for the role and listen for specifics instead of judging technical depth yourself.