CVS Health · Primly Community

CVS Health onsite / final round interview: how it really goes in 2026

market_realist · 4 replies

did the CVS Health final round (they call it the 'onsite' but it's fully virtual in 2026) last month for a senior data engineer role on their pharmacy analytics team. laying out the structure because i genuinely couldn't find this before going in.

format: 4 rounds back-to-back over about 3.5 hours, with two short breaks. virtual on Zoom with a shared screen. one Miro board for the design round.

round 1: technical (60 min). python and SQL coding. first problem was a data transformation task, medium difficulty, basically something you'd write in production. second was a SQL optimization problem against a healthcare-ish schema, patients and prescriptions. nothing leetcode-hard, just clear and practical.

round 2: system/data design (45 min). given a rough spec and asked to design a data pipeline for aggregating prescription fill data across regions with a focus on latency and data quality. interviewer was senior, asked follow-up questions. spent about 10 minutes on data modeling, then moved into infra choices, then error handling. they cared a lot about what happens when upstream data is late or malformed.

round 3: behavioral (45 min). two interviewers, rotating questions. all STAR. focus was on cross-team work and handling ambiguity. specific questions included "tell me about a time stakeholders had conflicting data needs" and "when have you had to push back on a deadline."

round 4: hiring manager (30 min). completely conversational. he asked about what i'm looking for, what energizes me at work, my take on the current team setup. i asked about the data stack (Snowflake, dbt, some Spark), the roadmap, and team size. no new technical content.

total debrief time: they said up to 5 business days. i heard back in 4.

4 replies

ds_dmitri

the 'late or malformed upstream data' question is a really good signal about what they actually deal with in production. healthcare data from external systems (pharmacy benefit managers, EHRs) is notoriously messy. that's not a hypothetical for them.

analyst_ana

did you feel like the snowflake/dbt stack was something they expected you to already know, or was it more about general data engineering principles?

de_derek

they expect you to know the concepts cold but weren't dogmatic about specific tool syntax. i've used Redshift and Databricks more than Snowflake and it wasn't a problem. the modeling and pipeline design thinking matters way more than tool familiarity.

careerveteran

4 rounds in 3.5 hours with breaks is pretty civilized for a virtual onsite. some companies stack 5-6 rounds without breaks and then wonder why candidates are fading by round 4.