Citi Data Engineer Interview Guide

Citi's data engineering loop is filtering for engineers who can operate within constraints without losing engineering judgment. The business runs regulatory reporting pipelines where a missed SLA carries legal consequence, not just an on-call page. What the interviewers are extracting is evidence that you understand compliance as an engineering input, not an afterthought: data lineage, auditability, and quality gates built into the pipeline design, not bolted on later. Citi's scale means you'll rarely own a greenfield problem; the signal they want is that you can inherit a Hadoop-era system, reason about its failure modes, and modernize it incrementally without breaking downstream consumers. Engineers who frame every answer around speed and autonomy tend to read as misaligned. The loop rewards candidates who treat organizational process as a constraint to engineer around, not a bureaucracy to complain about.

Start prep with SQL, because the screen will cover it and a weak SQL showing ends the process before the architecture rounds matter. 2 to 3 days of window functions, aggregations on financial-style schemas, and data quality query patterns (finding duplicates, detecting gaps in time series) puts you in range. Then shift to pipeline architecture: practice designing Spark batch jobs with explicit answers for late data, backfill strategy, and downstream impact. Don't spend time on streaming frameworks unless you can confirm the team you're interviewing for runs them; most Citi DE work is batch. The senior bar at L5 requires you to own the design end-to-end and anticipate regulatory implications without prompting; at mid-level, demonstrating solid execution and awareness of failure modes is enough. Skip prepping for greenfield lakehouse or ML platform questions; the hiring volume at Tampa and the stated focus on pipeline architecture both point toward infrastructure modernization work, not new platform builds.