Optiver's Data Scientist interviews combine a timed online assessment, probability and Fermi-estimation puzzles, and a take-home or data challenge. One candidate reported live trading games, and another faced tough questions on model selection and regularization.
$140K
Avg. Base Comp
$185K
Avg. Total Comp
4-5 rounds
Typical Rounds
1-2 months
Process Length
Optiver's data scientist interviews lean hard into quantitative reasoning rather than conventional ML trivia. Across four candidate accounts, the process opens with a timed online assessment that blends coding, probability, math games, and Fermi-style estimation. Two reports describe it as lasting two to three hours.
One candidate described a coding section built around a random-walk card game, where the hint pointed toward recursion. It also included a streaming trade-analysis problem that asked for the worst trade per instrument. Another candidate recalled binary-sequence coding and a lightbulbs puzzle. Candidates repeatedly flag the online assessment as heavily time-pressured, and one account ended at this stage.
Once through the OA, an HR or behavioral screen typically follows, focused on motivations, past mistakes, and teamwork. One candidate read the teamwork emphasis as a signal that collaboration matters at Optiver. The technical rounds that follow vary by account. One candidate reports live trading games that escalated in complexity from round to round. Others describe Fermi problems, game-theory questions, a coin-toss payoff game, Markov chains, or the three-legged-table puzzle.
A distinguishing feature is the take-home or data-challenge component. In one account, the results presentation drew probing questions on model selection:
That candidate's process ended there despite feeling confident earlier.
Given the small sample, treat round counts and total timeline as ranges rather than a fixed script. Expect quantitative reasoning and the ability to justify modeling choices to matter as much as raw coding speed.
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 Optiver process.
The process started with an online assessment on HackerRank that felt like a marathon—three hours covering coding, probability, statistics, and numerical sequences all in one go. I remember one question asked me to code up a trade analyzer from an order book, which combined algorithmic thinking with domain knowledge. After that initial filter, I moved to an HR phone screen that was pretty straightforward, mostly behavioral questions about mistakes I'd made and motivations for the role.
The technical interview came next, and this is where things got intense. I was hit with Fermi problems and game theory questions right away, which tested intuition more than pure coding. Then the conversation shifted to probability and statistics—they asked about a table with three legs placed randomly on a circle and whether it would fall, and probed deeper into Markov chains and coin toss scenarios. When we got into the data challenge portion, I had to present my results, and that's where I stumbled. The interviewer went deep on feature correlations, asking why I'd chosen one model over another and justifying specific hyperparameters. They pushed hard on regularization choices, specifically ridge versus lasso, and kept asking me to defend my assumptions about supervised versus unsupervised approaches.
The whole thing stretched over about a month with a fairly casual vibe, but the time pressure was real—especially in the online assessment and during that technical round. I didn't make it past the data challenge presentation, which honestly surprised me since I'd felt okay about the earlier rounds. Looking back, I think I underestimated how thoroughly they'd drill into the why behind every modeling choice.
Prep tip from this candidate
Expect intense questioning on model selection rationale and regularization trade-offs (ridge vs. lasso, supervised vs. unsupervised). Practice Fermi problems and probability questions like the three-legged table scenario, and be prepared to present and defend a data challenge result under scrutiny—that presentation round is critical and high-pressure.
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 Optiver
Write a function compute_deviation that returns the standard deviation of each list in a list of dictionaries
| Question | |
|---|---|
| Experiment Validity | |
| 500 Cards | |
| Session Difference | |
| Maximum Profit | |
| Sum to N | |
| Network Experiment Design | |
| Profit-Maximizing Dice Game | |
| Scalped Ticket | |
| Flipping 576 Times | |
| Stranded Miner | |
| Hurdles In Data Projects | |
| Priority Queue Using Linked List | |
| 85% vs 82% | |
| New UI Effect | |
| Bootstrapping Confidence Intervals | |
| Digital Marketing Metrics | |
| Interquartile Distance | |
| Slow SQL Query | |
| Addressing Data Quality Issues | |
| Index Fund Return | |
| Minimum Days for Scheduling All Meetings | |
| Client Solution Pushback | |
| Ranking Metrics | |
| Identifying Good Investors | |
| Processing Large CSV | |
| LRU Cache 1 | |
| Six Face Die | |
| 2nd Highest Salary | |
| Empty Neighborhoods |
Synthesized from candidate reports. Individual experiences may vary.
Candidates report a timed online assessment mixing coding problems, probability and math puzzles, and Fermi-style estimation questions. Two accounts describe it as lasting two to three hours. One account described a random-walk card game, where the hint pointed toward recursion, and a streaming trade-analysis problem that asked for the worst trade per instrument. Another recalled binary-sequence coding and a lightbulbs puzzle. Multiple accounts describe this stage as heavily time-pressured, and one candidate's process ended here.
After the OA, candidates typically move to an HR phone screen focused on behavioral questions: motivations for joining Optiver, past mistakes, and general fit. One candidate noted this round emphasized teamwork, which they read as a signal that collaboration is valued. Accounts describe this stage as fairly straightforward.
Technical rounds vary by account. Some candidates describe Fermi estimation and game-theory questions, followed by probability puzzles such as a three-legged table placed randomly on a circle, coin-toss games, and Markov chains. One candidate reports live trading games that escalated in complexity across rounds, with more variables and more participants.
Several candidates describe a take-home assignment or data challenge. In one account, the results presentation drew probing questions on model-selection rationale, hyperparameter choices, ridge versus lasso, and supervised-versus-unsupervised assumptions, and that candidate's process ended at this stage. Another candidate described a later discussion of ML techniques and their assumptions that felt relatively relaxed.
For some candidates, the process ends with a longer final stage. One account describes a roughly five-hour onsite with multiple interviewers cycling through. Another describes final interviews split between HR and a technical discussion of the take-home assignment. Reported end-to-end timelines range from about one month to roughly two months.
