Cisco · Primly Community

Cisco data engineer interview: pipelines and SQL were the whole game

de_derek · 4 replies

Just wrapped my Cisco data engineer interview loop in San Jose. Here is the breakdown because I couldn't find anything specific when I was prepping.

Phone screen with a recruiter, then a 45-min technical with a DE on the team. The technical was almost entirely SQL. Not toy queries, actual multi-join aggregations, window functions, ranking. They gave me a schema representing something like a network event log and asked me to find anomalies by device. If you're coming from a warehouse background, think Redshift or Snowflake dialect. They weren't picky about syntax but cared about whether you understood partitioning and indexing.

Onsite was four rounds: SQL + data modeling: schema design for a telemetry ingestion pipeline. I drew out a star schema. Interviewer asked about tradeoffs vs. wide flat tables. Pipeline architecture: walked through how I'd build an end-to-end Kafka to Spark Streaming to Postgres pipeline for device health metrics. They care about fault tolerance, exactly-once semantics, late-arriving events. Python / coding: one medium-difficulty problem, something like transforming nested JSON event records into a flat tabular format. Reasonably straightforward if you do this day-to-day. Behavioral: standard STAR stuff. Conflict on the team, a time you owned a project end to end, dealing with ambiguous requirements.

I noticed they really care about data quality. Multiple interviewers asked about how I'd detect upstream schema drift and what happens when source data is late or duplicated. Have a real answer for this.

Level they were hiring for was roughly what I'd call a mid-senior DE. Not entry. You need to have actually run a pipeline in production and have stories about it.

Timeline: phone screen to offer was about 5 weeks, which felt long but they were upfront about it. No lowball on comp, was roughly in line with San Jose market for the level.

Happy to answer specifics if anyone has questions about the SQL round.

4 replies

analyst_ana

this is incredibly helpful. did the SQL round feel more like leetcode-style questions or more like 'real' analysis queries? asking because i'm coming from a BI/analytics background and mostly write reporting SQL, not LeetCode-style hard stuff.

de_derek

more like real analysis queries, honestly. the hard part was less 'solve this puzzle' and more 'walk me through your reasoning as you go'. the window function stuff was ranking + lag/lead which is pretty standard analytics SQL. if you can write a rolling 7-day average over a partitioned dataset you're probably fine.

ds_dmitri

schema drift question is a really good signal for what they care about. i got asked something similar when i interviewed at a networking infra company a while back. having a concrete answer about data contracts, alerting on column nullability changes, that kind of thing goes a long way.

sre_sol

5 weeks is pretty normal for Cisco in my experience. they're not a startup. if you're on a 60-day visa clock that timeline gets stressful though.