JPMorgan Chase · Primly Community

JPMorgan Chase data engineer interview, pipelines and SQL: what they actually tested

de_derek · 4 replies

Went through the JPMC data engineer interview loop in March 2026 for a role on the commercial banking data platform team in NYC. Sharing details because I found almost nothing useful online before going in.

The process was five rounds total. Phone screen with a recruiter (comp expectations, timeline, standard stuff). Then a technical screen with a senior DE on the team. Then three onsite rounds back to back over two days, which they did virtually.

Technical screen was 60 minutes. Half SQL, half "walk me through a pipeline you built and what broke." The SQL was legitimately hard. Window functions, CTEs, a question that involved ranking with ties. Not medium LeetCode, this was actual query optimization thinking. I had to explain why I'd choose a certain join order and what the query planner would do.

Onsite round 1: system design for data. They gave me a scenario: design the ingestion layer for real-time transaction data feeding a fraud detection model. I talked through Kafka, the partitioning strategy, schema registry, how you handle late-arriving data. They pushed back hard on my retry logic. Be ready to defend your choices.

Onsite round 2: more SQL and Python. SQL again was window functions. Python was pandas, then they pivoted to "how would you do this in PySpark" for one of the transformations. I hadn't used Spark heavily in my current job and it showed a bit.

Onsite round 3: behavioral. Standard JPMC behavioral format. Three STAR-method questions. "Tell me about a time you caught a data quality issue in production." "Describe a time you had to influence a stakeholder without authority." Classic stuff but they want specifics.

A few things that mattered more than I expected: understanding data lineage, knowing the difference between batch and streaming architectures beyond the surface level, and being able to talk about observability on your pipelines (not just whether they run, but how you know the data is right).

Total timeline from application to offer was about 6 weeks. Offer came in at a reasonable senior DE band for NYC though I'll note JPMC is not trying to match pure tech company comp. If you need to know exact numbers, check the comp thread on this company page.

4 replies

ds_dmitri

The SQL bar sounds legitimately high. Was it more analytical (aggregations, window functions over time-series data) or more schema-design and normalization type questions? Asking because I'm interviewing for a data scientist role there and not sure how much to expect these kinds of questions to bleed over.

de_derek

Mostly analytical. Aggregations, window functions, ranking, one CTE-heavy query. Normalization came up only briefly in the system design context. For a DS role I'd guess the SQL is similar but the system design round would be more ML-pipeline oriented.

backend_bekah

The real-time fraud detection design question is basically a rite of passage at every fintech and bank I've interviewed at. Almost verbatim. Good idea to have Kafka + schema registry in your back pocket, also know what exactly-once semantics means and when you actually need it vs. when it's overkill.

sre_sol

Pipeline observability being a real topic in interviews is something I've noticed more across the board in 2026. It used to be "does the DAG succeed", now interviewers want to know how you detect data drift and silent failures. Good heads up.