Perplexity · Primly Community

Perplexity data engineer interview: pipelines and SQL were the whole deal

ds_dmitri · 6 replies

just finished my Perplexity data engineer loop a couple weeks ago. sharing what i know because there wasn't much out there when i was prepping.

recruiter screen was quick. maybe 25 minutes. she confirmed they use Spark, Airflow, and Snowflake and said to expect heavy SQL and some system design for data pipelines. that was accurate.

technical phone screen (45 min) first hour was one interviewer, very SQL-heavy. window functions, CTEs, a multi-step aggregation where you had to compute rolling stats on a time series. not toy queries, the kind where you're actually thinking about how many table scans you're doing. they seemed to care about correctness first, then efficiency.

onsite (virtual, 4 rounds) pipeline design: design an ingestion pipeline for high-volume search query logs. think about latency, deduplication, schema evolution. i drew out a streaming vs batch tradeoff. no single right answer, they pushed back to see how you reasoned. SQL deep dive: two more complex queries. one involved sessionization from a raw event table. classic. coding: one Python problem, not leetcode-style, more like "parse this log format and compute X". practical. cross-functional: talked with someone from the ML platform team about how data eng hands off to ML. they wanted to know if i'd worked at the intersection.

what mattered: they kept asking about scale. not FAANG scale necessarily, but what happens when your pipeline assumptions break. had they asked me to code a perfect solution or asked me to tell them where it breaks? the second one.

no take-home. total elapsed time: 3 weeks from first call to verbal offer.

my background: 7 YOE, mostly fintech data platform work. leveled as senior DE.

happy to answer questions.

6 replies

ae_andre

really useful, thanks. how specific did they get on Airflow? like DAG design patterns or just general familiarity?

de_derek

more conceptual than deep. i talked about dynamic DAG generation and idempotent tasks and they seemed satisfied. they weren't quizzing on Airflow internals.

backend_bekah

the sessionization SQL problem is everywhere right now. every data shop i've interviewed at this year has some variant of it.

visa_vik

did they ask about H1B sponsorship early? i'm trying to figure out if it's worth applying as someone who needs sponsorship.

de_derek

i'm a citizen so i can't speak to that personally. my recruiter didn't mention it either way. probably worth asking directly on the first call before spending 4+ rounds.

analyst_ana

this is the most detailed Perplexity DE breakdown i've seen. bookmarking before my loop next month.