Plaid · Primly Community

Plaid data engineer interview, pipelines and SQL and what else they actually test

ds_dmitri · 4 replies

Completed the Plaid data engineering loop a couple months ago for a senior DE role. Sharing because data eng interviews vary wildly by company and I spent two weeks prepping for the wrong things.

Here's the actual format: one OA (SQL-heavy), one live coding round, one system design round focused on data pipelines, one behavioral. Four rounds plus the recruiter screen.

The OA SQL: harder than I expected. Not just joins and group-bys. They had a question involving window functions with partitioning, and another that was basically a recursive CTE or a creative use of LAG/LEAD to solve a gap-and-island type problem. If you've been lazy about advanced SQL, sharpen up before applying. I'd say the SQL OA is medium-hard LeetCode equivalent if you calibrate against the SQL section there.

Live coding: Python. A data processing problem. They gave me a dataset-like structure and asked me to transform it, handle edge cases, and then optimize for larger input. Felt like a real data task, not a pure algorithms problem. Understanding pandas would help but they let me use pure Python if I preferred.

System design (the important one): design a financial data pipeline. The context was very Plaid-specific: imagine you're ingesting real-time transaction data from bank connections at scale, normalizing it across inconsistent bank response formats, and serving it reliably to downstream applications. What does that architecture look like?

I talked through: ingestion layer (Kafka for streaming), normalization service (schema registry, handling missing fields), storage (data lake + hot cache for recent transactions), retry logic for failed ingestion, observability. The interviewer pushed hard on "what happens when a bank returns malformed data" and "how do you handle schema drift over time."

Financial data note: they care about auditability. Every pipeline change should be traceable. I mentioned immutable event logs and that landed well.

Total loop: about 5 weeks start to finish. Not lightning fast but not slow. They were communicative.

Comp for senior DE in SF: my offer was around $220-240k TC, mostly salary-heavy with a modest equity component compared to pure SWE roles. Worth asking about equity vesting schedule specifically.

4 replies

ops_omar

The gap-and-island SQL problem on an OA is kind of brutal. That's usually a "do you actually understand window functions" filter. Did they give you any hints or was it fully cold?

analyst_ana

Do they hire more junior data engineers or is the DE loop pretty much all senior-level? I'm a BI analyst with 3 years of experience thinking about moving into DE and wondering if Plaid would even consider me.

de_derek

I only saw senior/staff DE roles listed when I was searching. That doesn't mean junior roles don't exist, just that they're not publicly posted at the same volume. Might be worth reaching out to a recruiter directly to ask. They seemed open to that kind of question.

infra_ines

Schema drift over time is such a real problem with external data sources. I'd be curious what answer they were looking for there. Did they want schema registry with versioning, or more like contract testing, or something else entirely?