Four candidates describe Optum Data Scientist interviews that mix project deep dives, ML fundamentals, and either coding, a take-home, or a team round. Here is what each reported and how to prepare.
$125K
Avg. Base Comp
$145K
Avg. Total Comp
3-4 rounds
Typical Rounds
2 months
Process Length
Candidates describe Optum Data Scientist interviews as built around your own past work. One candidate was asked to walk through shipped projects across several rounds and, in at least one, to present an ML model. Another reports a resume-heavy technical discussion, and a third reports a four-round process where the second technical round dug into ML and deep learning projects. Be ready to explain why you made each modeling choice, not only what you built.
The formats differ by candidate, so prepare for more than one. One reports a two-hour online exam with three data structures and algorithms questions and three SQL questions at medium difficulty. Another reports a live coding task covering preprocessing, missing values, label encoding, model training and F1 evaluation. A third reports a recruiter screen, a take-home analysis of complex medical data aimed at finding cost-saving opportunities, a presentation to two data scientists, and a manager conversation on behavior and fit.
Technical topics reported include random forest and XGBoost internals, handling a dataset with poor fill rates, boosting trade-offs, class imbalance, LoRA and QLoRA, cloud platforms, Python libraries and MongoDB. Practice explaining these out loud with examples from your own projects.
Two candidates report that the process took about two months. One of them attributes part of that to hiring freezes. Round counts reported were three and four, with HR or director conversations near the end. Three of the four candidates report receiving an offer, and one reports no offer after the technical rounds.
Synthesized from 4 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 Optum process.
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 Optum
Select the 2nd highest salary in the engineering department
| Question | |
|---|---|
| Size of Joins | |
| String Mapping | |
| Sort Strings | |
| Hurdles In Data Projects | |
| Concurrent LLM Serving | |
| String Palindromes | |
| Impossibly Iterative Fibonacci | |
| Azure Kubernetes Infrastructure | |
| Account Personalization Strategy | |
| Your Strengths and Weaknesses | |
| Fast Food Database | |
| Experiment Validity | |
| Prime to N | |
| Liked Pages | |
| Weighted Keys | |
| Button AB Test | |
| Decreasing Comments | |
| P-value to a Layman | |
| Network Experiment Design | |
| Bagging vs Boosting | |
| Daily Retention Summary | |
| Delivery Estimate Model | |
| Detecting ECG Tachycardia Runs | |
| The Brackets Problem | |
| Google Maps Improvement | |
| Declining Applicants | |
| Random Forest Explanation | |
| Subscription Retention | |
| Valid Anagram |
Synthesized from candidate reports. Individual experiences may vary.
One candidate reports applying through the online portal, and another reports a recruiter screen focused on motivations. Recruiters were described as communicative and flexible on scheduling. Expect a short conversation about your background and why you want the role, though the exact format may vary by team.
Candidates report different first technical hurdles. One describes a two-hour online exam with three data structures and algorithms questions and three SQL questions at medium difficulty. Another describes a live task covering preprocessing, missing values, label encoding, training a model and evaluating it with F1. Not every candidate reported this stage.
Several candidates report technical conversations driven by their resume and past projects. Topics reported include random forest and XGBoost internals, missing-data strategy, boosting trade-offs, class imbalance, LoRA and QLoRA, cloud platforms and Python libraries. One candidate presented an ML model they had built. Interviewers were described as technical and conversational.
One candidate reports a take-home test using complex but public medical data, where the task was to find opportunities to reduce cost and make recommendations. Two data scientists then reviewed the work and asked questions. The candidate found nothing unfair or unexpected. Other candidates did not report this stage.
Final stages reported include a behavioral round on problem-solving and collaboration, an in-person session with the team, a manager conversation on behavior and fit, and director and HR rounds. One reported question was how you handle meetings that go off topic. The exact order and format likely vary by team.
