Flatiron Health Data Scientist candidates report a recruiter call, timed or live coding in SQL, Python, pandas or R, a project deep dive, and ML metrics questions such as precision and recall.
$151K
Avg. Base Comp
$223K
Avg. Total Comp
2-5 rounds
Typical Rounds
Not reported
Process Length
Candidates describe Flatiron Health's Data Scientist interviews as fairly approachable technically, but the format varies from person to person. The most repeated element is a walkthrough of a past project: four of the five reports mention being asked to explain their background or an impactful project, and one candidate described a full hour on the choices they made and what they would change.
Coding is the other constant, though the tools differ. Reports mention a HackerRank screen, pandas manipulation, live Python questions, SQL and R through a Karat session, and a timed 60-90 minute SQL and analytics exercise. One candidate said the SQL was moderately difficult, with joins and aggregations. Another found the coding easy but still a test of comfort under pressure. Ask your recruiter exactly what the technical rounds include, because one candidate was surprised by a 30-minute live coding session midway through an interview.
Machine learning questions look conceptual rather than exotic. Two candidates were asked to compute or explain precision and recall from a confusion matrix, and one was asked to diagnose why a model's classification performance was low. One candidate who withdrew before finishing said the full loop also included a data modeling component, with questions about designing tables to support particular analytics queries.
None of the five reports ended in an offer, and one of those candidates withdrew, so treat the rounds as a sample rather than a fixed loop. Two of the reports are from senior-level applicants. Prepare SQL, one of R or pandas, metric definitions and a clear project story.
Synthesized from 5 candidate reports by our editorial team.
Had an interview recently?
Share your experience. Unlock the full guide.
Real interview reports from people who went through the Flatiron Health process.
I applied for a Senior Data Scientist role with about 8 years of machine learning experience in healthcare, including work with Flatiron data, so I expected the process to focus on my background. The first round was a 30-minute phone call with talent acquisition. The second round was a one-hour technical deep dive, and partway through the interviewer told me I would also be doing 30 minutes of live coding in SQL and Python. Nobody had mentioned this beforehand, and I had not prepared for it.
The coding exercise did not seem connected to the work I could see Flatiron doing. When I said it was irrelevant to my experience and to the role, the explanation I got was that it was designed to give candidates without healthcare experience a chance. I did not find that convincing, and I said so.
The rest of the hour did not go the way I expected either. The main question was to describe a project I had worked on. There was no real technical deep dive into the methods I had used, and I got little sense that the interviewer knew ML or data science well enough to assess the depth of my experience. From my side, the interview was not especially challenging. It was mostly confusing.
Two days later I received an automated rejection email with no feedback. Looking back, the biggest lesson for me is that a role's assessment format can change without much warning, and it is worth pinning down early what the technical rounds will include.
Prep tip from this candidate
Ask the recruiter before the technical round whether any live coding is included and what it covers, since this candidate only learned about a 30-minute SQL and Python session mid-interview. Be ready to walk through a past project end to end, because that was the main question asked.
Share your own interview experience to unlock all reports, or subscribe for full access.
Sourced from candidate reports and verified by our team.
Topics based on recent interview experiences.
Featured question at Flatiron Health
Write a function to calculate precision and recall metrics.
| Question | |
|---|---|
| Book Combinations | |
| 2nd Highest Salary | |
| Top Three Salaries | |
| Experiment Validity | |
| Top 3 Users | |
| Prime to N | |
| Weighted Keys | |
| Button AB Test | |
| Size of Joins | |
| Hurdles In Data Projects | |
| Network Experiment Design | |
| Bagging vs Boosting | |
| Google Maps Improvement | |
| Detecting ECG Tachycardia Runs | |
| Type I and II Errors | |
| The Brackets Problem | |
| P-value to a Layman | |
| Ride-Sharing App Schema | |
| Skewed Pricing | |
| Declining Applicants | |
| Random Forest Explanation | |
| Justify a Neural Network | |
| String Mapping | |
| Sort Strings | |
| Valid Anagram | |
| Xgboost vs Random Forest | |
| Find Duplicate Numbers in a List | |
| Swap Variables | |
| Measuring Customer Service Quality |
Synthesized from candidate reports. Individual experiences may vary.
Candidates report an initial recruiter conversation, in one case about 30 minutes, that outlined the interview steps in advance. One candidate instead received a timed technical screen by email with no preliminary call. Expect to cover background and availability, and confirm which technical formats come next.
Candidates report different first technical screens: a HackerRank test, a one-hour Karat session in SQL and R, and a 60-90 minute SQL and data analytics exercise. Tasks were described as basic to moderately difficult, covering joins, aggregations and data manipulation. The exact format may vary by role or timing.
Candidates describe an hour-long conversation with a hiring manager or team member that combined project discussion with live coding in Python, pandas, R or SQL. Reported questions included precision and recall from a confusion matrix, motivation and background, and one candidate was surprised by a 30-minute live coding session.
One candidate described a three-hour super day with a project deep dive, live R coding and a model diagnosis scenario. Another reported final conversations of about an hour each with team members. A withdrawn candidate was told the full loop could include case studies, behavioral, coding and data modeling rounds, so later stages may differ.
