Databricks interview questions: answers checked against the docs
Lakehouse questions for data engineering and analytics roles, answered in plain words and checked against the Databricks documentation.
Databricks interviews cover Delta Lake and its transaction log, the medallion architecture, Unity Catalog, compute choices, time travel, OPTIMIZE and VACUUM, MERGE for upserts, Auto Loader, schema enforcement and evolution, liquid clustering, Photon, change data feed, and orchestration with jobs and declarative pipelines. Below are 13 questions; each answer lists the Databricks documentation pages it was checked against.
Basic Databricks interview questions
Basic1. What is Delta Lake, and why does Databricks use it?
Delta Lake stores table data as Parquet files plus a transaction log that records every change. The log gives tables ACID transactions (readers never see half-written data, concurrent writers do not corrupt each other), table versions and scalable metadata on top of cheap object storage. On Databricks it is the default table format.
Checked against: Databricks docs: Delta Lake
Basic2. What is the medallion architecture?
A way of organizing a lakehouse in layers by data quality: bronze holds raw data as ingested, silver holds cleaned, validated and conformed data, and gold holds business-level aggregates ready for reporting. It is a design pattern, not a product feature, and teams adapt the number and naming of layers to their needs.
Checked against: Databricks docs: medallion architecture
Basic3. What is Unity Catalog?
Databricks' governance layer: one place to manage permissions, auditing and lineage for tables, views, volumes, functions and models across workspaces. Objects use a three-level name, catalog.schema.table. Managed tables let Unity Catalog handle both governance and storage, while external tables keep data in a location you manage.
SQL snippet · illustrative, not run on this pageSELECT * FROM main.sales.orders;
GRANT SELECT ON TABLE main.sales.orders TO `analysts`;Checked against: Databricks docs: Unity Catalog
Basic4. What compute options does Databricks offer?
Serverless compute, which Databricks provisions and scales for you; classic compute (clusters you configure yourself, either all-purpose for interactive work or job compute for scheduled runs); and SQL warehouses for SQL queries and BI tools. In an interview, connect the choice to the workload: interactive exploration, a scheduled pipeline or dashboard queries.
Checked against: Databricks docs: compute, Databricks docs: SQL warehouses
Intermediate Databricks interview questions
Intermediate5. What is time travel, and what limits it?
Every write creates a new table version, so you can query an earlier version or timestamp, audit changes with DESCRIBE HISTORY, or restore a table after a bad write. Time travel depends on old data files still existing: VACUUM deletes unreferenced files older than its retention threshold (7 days by default), after which those versions can no longer be queried. Table history is not a backup.
SQL snippet · illustrative, not run on this pageDESCRIBE HISTORY main.sales.orders;
SELECT * FROM main.sales.orders VERSION AS OF 12;
RESTORE TABLE main.sales.orders TO VERSION AS OF 12;Checked against: Databricks docs: table history, Databricks docs: VACUUM
Intermediate6. What do OPTIMIZE and VACUUM do?
OPTIMIZE rewrites many small files into fewer, larger ones (and applies clustering), which speeds up reads. VACUUM removes data files that the table no longer references and that are older than the retention threshold, which saves storage. They solve different problems: OPTIMIZE is about performance, VACUUM about cleanup, and VACUUM shortens how far back time travel can go.
SQL snippet · illustrative, not run on this pageOPTIMIZE main.sales.orders;
VACUUM main.sales.orders; -- default retention: 7 daysChecked against: Databricks docs: OPTIMIZE, Databricks docs: VACUUM
Intermediate7. How do you upsert data into a Delta table?
With MERGE INTO: match source rows to target rows on a key, update the matches, insert the rest and optionally delete. It is the standard way to apply change data or deduplicate incoming records. Make sure the source has one row per key: the merge can fail when several source rows match and try to update the same target row.
SQL snippet · illustrative, not run on this pageMERGE INTO main.sales.customers AS t
USING updates AS s
ON t.customer_id = s.customer_id
WHEN MATCHED THEN UPDATE SET *
WHEN NOT MATCHED THEN INSERT *;Checked against: Databricks docs: MERGE
Intermediate8. How does Auto Loader ingest files incrementally?
Auto Loader is a Structured Streaming source (cloudFiles) that picks up new files as they land in cloud storage. It records which files it has processed in the stream's checkpoint, so a restart continues where it left off and files are not loaded twice. It can also infer the schema and handle new columns.
PySpark snippet · illustrative, not run on this page(spark.readStream.format("cloudFiles")
.option("cloudFiles.format", "json")
.option("cloudFiles.schemaLocation", "/Volumes/main/raw/_schemas/orders")
.load("/Volumes/main/raw/orders/")
.writeStream
.option("checkpointLocation", "/Volumes/main/raw/_checkpoints/orders")
.toTable("main.bronze.orders"))Checked against: Databricks docs: Auto Loader
Intermediate9. What is schema enforcement, and how do you evolve a schema?
Delta tables reject writes whose columns do not match the table's schema, which stops bad data from silently landing. When a change is intended, such as a new column, you opt in to schema evolution for that write with mergeSchema (or WITH SCHEMA EVOLUTION in SQL).
PySpark snippet · illustrative, not run on this page(new_orders.write
.mode("append")
.option("mergeSchema", "true")
.saveAsTable("main.silver.orders"))Checked against: Databricks docs: schema evolution
Advanced Databricks interview questions
Advanced10. What is liquid clustering, and how does it compare with partitioning and Z-ordering?
Liquid clustering organizes data by clustering keys to speed up filters, and replaces both Hive-style partitioning and ZORDER for new tables. Its main advantage is that you can change the clustering keys later without rewriting existing data, whereas changing a partition column means rebuilding the table. Partitioning is still seen on older tables, so know both.
SQL snippet · illustrative, not run on this pageCREATE TABLE main.sales.events (event_date DATE, customer_id BIGINT, action STRING)
CLUSTER BY (event_date, customer_id);
ALTER TABLE main.sales.events CLUSTER BY (customer_id);Checked against: Databricks docs: liquid clustering
Advanced11. What is Photon?
Photon is Databricks' native, vectorized query engine: it processes data in columnar batches and speeds up SQL and DataFrame workloads. It is compatible with Spark APIs, so existing code runs unchanged; whether it helps depends on the workload, so compare run time and cost before and after enabling it.
Checked against: Databricks docs: Photon
Advanced12. What is change data feed, and when would you use it?
When enabled on a table, change data feed records row-level inserts, updates and deletes between versions. Downstream jobs can read only what changed since their last run instead of rescanning the table, which makes it a common building block for incremental silver-to-gold processing and for syncing to other systems.
SQL snippet · illustrative, not run on this pageALTER TABLE main.silver.orders SET TBLPROPERTIES (delta.enableChangeDataFeed = true);
SELECT * FROM table_changes('main.silver.orders', 5);Checked against: Databricks docs: change data feed
Advanced13. How do you orchestrate and build pipelines on Databricks?
Lakeflow Jobs schedules and runs tasks such as notebooks, SQL and Python scripts with dependencies, retries and alerts. For the transformations themselves, Lakeflow pipelines let you declare streaming tables and materialized views in SQL or Python while the platform manages execution and dependencies. Be ready to say when you would pick a declarative pipeline over hand-written jobs.
Checked against: Databricks docs: Lakeflow pipelines, Databricks docs: Lakeflow Jobs
How to practice with this page
- Walk through one pipeline end to end: ingest, bronze, silver, gold, and how each step is scheduled.
- Explain each maintenance command by what it changes in storage.
- Write a MERGE and a time travel query by hand and explain what each clause does.
- Review Spark fundamentals: partitions, shuffles, joins and skew.
Related interview prep and tools
The SQL that Databricks roles test.
Open →QuestionsPython interview questionsFor PySpark and notebook work.
Open →QuestionsData science interview questionsStatistics and modeling questions.
Open →QuestionsKafka interview questionsStreaming sources, checked against the docs.
Open →Free toolResume summary generatorDraft a summary for a Data Engineer role; the title is filled in, you add your experience.
Open →Free toolLinkedIn headline generatorWrite a headline for a Data Engineer search.
Open →Disclosure: TryHired makes HIRED, a paid job-search program, so we have a conflict of interest when we mention it. This question bank is free, with no TryHired signup.
Questions people ask
What Databricks topics come up often in interviews?
Do I need to know Apache Spark for a Databricks interview?
Were these snippets run?
How this page is built
TryHired wrote these questions and answers. Statements about the technology were checked against the official documentation pages listed below, saved when the page was built (2026-11-15). The code snippets are short illustrations and were not run on this page. The question list is general practice: it does not predict what a particular employer will ask.
- Interviewers ask different questions for different roles and levels.
- Answers are short explanations, not complete references.
- No Databricks workspace was used on the build machine, so nothing was run on this page; product names follow the documentation as read and change over time.
- Databricks docs: Auto Loader (read 2026-11-15)
- Databricks docs: change data feed (read 2026-11-15)
- Databricks docs: compute (read 2026-11-15)
- Databricks docs: Delta Lake (read 2026-11-15)
- Databricks docs: Lakeflow pipelines (read 2026-11-15)
- Databricks docs: Lakeflow Jobs (read 2026-11-15)
- Databricks docs: liquid clustering (read 2026-11-15)
- Databricks docs: medallion architecture (read 2026-11-15)
- Databricks docs: MERGE (read 2026-11-15)
- Databricks docs: OPTIMIZE (read 2026-11-15)
- Databricks docs: Photon (read 2026-11-15)
- Databricks docs: schema evolution (read 2026-11-15)
- Databricks docs: table history (read 2026-11-15)
- Databricks docs: Unity Catalog (read 2026-11-15)
- Databricks docs: VACUUM (read 2026-11-15)
- Databricks docs: SQL warehouses (read 2026-11-15)