Paramount · Primly Community

Paramount onsite / final round, how it really goes (data engineering perspective)

analyst_ana · 6 replies

did the full Paramount onsite for a senior data engineer role, NYC hybrid, early 2026. this is the complete picture.

format: 4 rounds back to back over one day (virtual). each 45-60 min with a break in the middle.

round 1: coding. medium SQL and Python combo. they gave me a dataset scenario (streaming engagement data, basically) and asked me to write a query to find the top 10 shows by watch time in the past 30 days, then pivot and add a windowing function. the Python part was more of a discussion about how i'd clean and process the same data in Spark. not a leetcode hard, more of a practical data eng question.

round 2: system design. for DE roles this was pipeline design, not backend infra design. they asked me to design the data pipeline for ingesting ad click events from multiple sources (third-party ad servers, internal apps) into a central warehouse for reporting. i walked through Kafka for ingestion, some light transformation layer, schema registry, and Snowflake as the warehouse. they pushed on exactly-once semantics and how i'd handle late-arriving events. deduplication window question, classic.

round 3: behavioral. 45 min, cross-functional stuff mostly. how i worked with analytics teams, a time i disagreed with a technical decision and what happened. pretty standard senior behavioral. see the other thread on this for more detail.

round 4: hiring manager conversation. this one was more casual, more of a two-way conversation. they explained the team's current stack (Airflow, Databricks, Snowflake, moving more things to real-time with Kafka). asked what i found most interesting about data engineering right now. i talked about streaming vs. batch tradeoffs and they got visibly engaged. probably the most enjoyable part.

timeline: onsite to offer was 11 business days. recruiter was good about keeping me updated.

vibe overall: collaborative, not adversarial. the interviewers genuinely seemed to be evaluating fit both ways. i felt like i could ask real questions and get real answers.

6 replies

pivot_pat

the late-arriving events question is such a consistent data eng probe. i've gotten it at like 4 companies. watermarking in Flink, dedup windows, it always comes up when streaming is part of the stack.

ml_mike

curious if ML roles have a similar onsite structure or if the system design round gets swapped out for something else. the recommendations team at Paramount+ does real work so i'd imagine their ML loop is more rigorous.

de_derek

i can't speak firsthand but someone from that team mentioned to me (after i joined) that ML roles have an ML system design round specifically, more focused on feature stores, offline/online training pipelines, serving infrastructure. the coding round stays algorithmic but the design round pivots to ML-specific scenarios.

staff_steph

the hiring manager round being a two-way conversation is a good sign culturally. when HM rounds are just another interrogation it usually means they haven't thought hard about what they're looking for. them actually explaining the stack and asking what you find interesting signals a functional team.

ops_omar

11 days from onsite to offer is actually fast. most of my onsite-to-offer timelines this year have been 3+ weeks. did they give you an exploding offer or was there time to consider?

de_derek

they gave me a week to decide which was fine. no pressure tactics. i asked for a couple extra days and they granted it without drama.