Sr. Data Scientist
Posted Sep 15, 2026 · We last checked this listing on Oct 6, 2026
Likely interview questions for this role
Written from this job description, not a generic list. Each one notes what the interviewer is really checking.
About your past
Walk me through a time you pulled data from several sources, like a database, flat files and an API, into one training set. What went wrong along the way?
What they are checking: Whether they have done the unglamorous data acquisition work themselves rather than receiving clean tables from someone else.
Describe a model you built that a business partner did not trust at first. How did you win them over, or did you?
What they are checking: Whether they can explain results plainly and accept that adoption, not accuracy alone, decides whether the work matters.
Tell me about a time the right answer was a simple model or no model at all.
What they are checking: Whether they put the business result ahead of showing off technique, which is a sign of senior judgment.
What would you do
You get a raw dataset with fields nobody has documented. How do you work out what each field really means before you model anything?
What they are checking: Whether they lean on descriptive statistics and domain experts instead of guessing, since wrong field meanings poison every model downstream.
Your model performs well in testing, but a month after launch the predictions start drifting. What do you check first?
What they are checking: Whether they have a calm, ordered way to separate data changes, pipeline bugs and real shifts in behavior.
In a regulated firm like ours, models often get reviewed by a validation team. How do you write and document your work so it survives that review?
What they are checking: Whether they expect scrutiny and build reproducibility and clear assumptions in from the start.
About the work
What unit tests and data quality checks do you put on a pipeline so a model never trains on bad data without anyone noticing?
What they are checking: Whether they treat trusted data as an engineering problem with automated checks, not a one-time manual look.
How do you split data into training, validation and test sets for a time series problem, and what mistake do people make there?
What they are checking: Whether they understand leakage from the future, which quietly inflates results on financial time series.
Tell me about an ensemble method you proposed. Why did it beat a single model, and how did you prove that?
What they are checking: Whether they can justify added model complexity with honest out-of-sample evidence rather than a leaderboard number.
When would you reach for an unsupervised approach instead of a supervised one on a business problem at a bank?
What they are checking: Whether they choose methods from the shape of the problem and available labels, not from habit.
How do you structure Python code for a data science project so someone else can rerun it six months from now?
What they are checking: Whether they write production-minded software, which the posting stresses, rather than notebooks only they can run.
Practice this interview out loud.
Offer builds a real interview for this exact role at Northern Trust from your resume and this job description, asks the questions one at a time, and tells you what landed. The first one is free.
Practice this out loudThe full job description
As published by Northern Trust.