Adobe · Primly Community

Adobe data engineer interview: pipelines and SQL, my full loop breakdown

de_derek · 4 replies

Just finished the Adobe data engineer interview loop for a role on the Digital Experience platform team. Sharing because I couldn't find much current info when I was prepping.

The recruiter screen was pretty standard. About 20 minutes, mostly clarifying what team I was applying to and confirming I'd done Spark/Airflow type work. Nothing technical.

First technical round: SQL. And I mean SQL. Not the "write a basic GROUP BY" kind. They gave me a realistic schema (something resembling event data from a web analytics product) and asked me to do multi-step aggregations, window functions (LAG, RANK, DENSE_RANK), and one query that required a self-join. About 60 minutes. I did it in a shared doc, not a real IDE, which adds pressure. Know your window functions cold.

Second technical round: pipeline design. This was a system design round but scoped tightly to data pipelines. They described an ingestion problem (high-volume clickstream events, some late-arriving data) and wanted me to walk through architecture choices. I talked through Kafka for ingest, Spark Structured Streaming vs. batch tradeoffs, partitioning strategy, and how I'd handle schema evolution. The interviewer pushed hard on the late data question. Know your watermarking story.

Third round: past project deep-dive. They picked one project off my resume and went three levels deep. How did you define success metrics, what broke first, what would you do differently. Pretty behavioral but engineering-flavored.

Fourth round: hiring manager. More of a fit/vision chat. She described the team's current stack (Databricks, some legacy Hadoop they're migrating away from, dbt on the transformation side) and asked what excited me about it.

Total time from recruiter screen to offer: about 5 weeks. The pipeline design round was the hard one for me. Prep that specifically.

4 replies

analyst_ana

This is so useful. Did they ask about specific Adobe products in the SQL round or was it a made-up schema? Also wondering if Python came up at all or if it was SQL-only.

de_derek

Made-up schema but very obviously inspired by web analytics events. Clicks, page views, sessions. Python came up briefly in the pipeline design round when I mentioned PySpark but nobody asked me to write any Python code.

ds_dmitri

The window function depth matches my experience at a similar Adobe adjacent role. LAG and DENSE_RANK are specifically worth drilling because people mess up the PARTITION BY ordering. Also: do you know what level the role was scoped at? Trying to calibrate.

infra_ines

5 weeks is actually fast for a company that size. My loop somewhere else dragged to 8 weeks and I had to cold-email the recruiter every two of those.