Bloomberg · Primly Community

Bloomberg data engineer interview, pipelines and SQL: what actually showed up

de_derek · 4 replies

Went through the Bloomberg data engineer loop in early 2026, NYC role. Sharing what actually came up because most of what I found searching was vague or three years old.

Phone screen was with a recruiter, standard background questions. Then a 45-min technical screen with an engineer. About half SQL, half Python data manipulation. The SQL was genuinely hard: CTEs, window functions, ranking, a couple of questions where you had to think through aggregation order to avoid double-counting. Not LeetCode-style puzzles, more like "here's a messy dataset, answer a business question." That framing felt very Bloomberg.

Onsite was four rounds over two days (virtual):

Data modeling / pipeline design. They gave a scenario around ingesting market data feeds: high volume, late-arriving data, deduplication. You're expected to talk schema choices, partitioning strategy, and how you'd handle reruns. Kafka came up. I mentioned exactly-once semantics and that generated a 20-minute side conversation, so know your streaming basics.

SQL depth round. More window functions. Also a question about query optimization, basically walk me through how you'd debug a slow query on a billion-row table. Execution plans, indexes, partition pruning.

Coding. Python. One string parsing problem, one that was essentially a join operation done in code (not SQL). Not hard on raw difficulty but time-pressured. Explain your reasoning the whole way.

Behavioral. Bloomberg's behavioral questions lean toward ownership and crisis response. "Tell me about a pipeline that failed in production and what you did." Prepare two or three real incidents. They don't want theoretical answers.

I came in with 7 years of data eng experience, mostly fintech, and the financial data angle helped. If you're from a different domain, learn a little about market data concepts: tick data, time-series, order books at a high level.

Total process was about 6 weeks from application to offer. Recruiter communication was decent, better than average actually.

4 replies

ds_dmitri

The double-counting aggregation trap is real. That's a classic Bloomberg SQL question pattern. I had the exact same type of question when I interviewed for a data analytics role there in 2024. Did they ask anything about Spark or was it mostly Python pandas-style?

de_derek

Spark came up in the pipeline design conversation but not as a hands-on coding question. I mentioned PySpark for the streaming scenario and they were happy with that. The coding rounds were pure Python, no distributed frameworks.

analyst_ana

Saving this. I have a Bloomberg data analyst screen next week and I'm worried the SQL bar is too high. Did they give you any prep materials or hint at what to study?

backend_bekah

The financial domain knowledge angle is underrated prep advice. Most DS/DE interview guides skip it entirely because they're written generically. For Bloomberg specifically it genuinely differentiates you.