
Company interview guide
Flow Traders Machine Learning Interview Guide (2026)
For ML, AI and research roles: a stats-and-ML or language-and-coding online test, a recruiter call, probability and regression interviews, a codepair for engineers, then a research deep dive with a team lead.
Flow Traders's machine learning interview loop has 8 reported stages, from application and resume screen to final round: research deep dive and team lead; typical length: Varies; neither the firm nor compiled accounts give a typical length.
Roles covered: Research Scientist, ML Engineer, Applied Scientist and Data Scientist. The question bank holds 6 questions reported from Flow Traders interviews, 2 of them free to solve in your browser.
- Interview process length
- Varies; neither the firm nor compiled accounts give a typical length
- Stages
- 8
- Reports linked to bank questions
- 6
- Bank report dates
- Jul 2026–Jul 2026
- Questions in bank
- 6 (2 free)
- Last reviewed
- October 2026
How the Flow Traders interview loop works
Apply with a CV and cover letter, then an online test: research applicants report two take-homes back to back (a 15-minute statistics and ML multiple-choice section, then 30 minutes of probability with one Python Monte Carlo task), while technology applicants take a test of language-quirk multiple choice plus three coding problems. Next comes a recruiter interview on motivation, business knowledge and fit, then technical interviews with the team: probability and brainteasers, statistics and regression depth, and for engineers a codepair with a senior developer. The final round is with a team lead or desk head and centres on your past research and fit; team allocation is decided during the interviews, and the firm books travel when finals are in person. The numerical mental-math and sequences test and the senior-trader case study are trader-track only, and there is no market-making game. The firm's pages put the test before the recruiter call, while compiled accounts list a recruiter screen first.
The Flow Traders interview process, stage by stage
01Application and resume screen
Online application with a CV and cover letter through the firm's job board; the recruitment team reviews every CV
Relevant openings include PhD Graduate Quantitative Researcher (Amsterdam, New York, Hong Kong), experienced and mid-frequency quant researcher roles, and Senior Research Engineer (London, Hong Kong), which leads the framework for defining, training and backtesting trading models. Graduate quant researchers get two months of training on pricing maths and automated strategies. The postings say a matching tool only highlights skills and people make every hiring decision.
What it tests
- Research: a PhD in maths, physics, computer science or similar, with strong statistics and linear algebra
- Implementing models and algorithms in Python or C++
- Research engineer: end-to-end ML pipelines with numpy, PyTorch, polars, pandas and Ray
- Ownership of open-ended problems
How to prepare
- Name the concrete stack you have used where it is true (PyTorch, polars, Ray, Docker, Kubernetes, CI/CD), since the research engineer posting lists it.
- Write the cover letter; the firm's pages ask for one on every track.
- Show one model that went from research into live use and how you monitored it.
The firm's careers pages and its quant researcher and research engineer postings describe the CV and cover-letter application; no first-hand ML-role report covers this step.
02Online assessment
Research track: two take-homes back to back, 30 questions in 15 min (25 multiple choice + 5 integral-heavy), then 30 min for 6 short-answer probability questions and 1 Python Monte Carlo task.
Technology track: HackerRank, about 7 language-quirk multiple-choice questions and 3 coding problems
The research test arrives as two links, the second right after you submit the first. Part one covers statistics concepts, ML (lasso and ridge, feature selection, sources of error), two or three Brownian motion questions and covariance matrices, plus five standalone integral questions, with no language-syntax items. Part two, labelled Fast Fire 2.2, asks for expected colour runs in a shuffled deck, a stopping problem with two consecutive misses, the mean distance between two random points in a unit square, a bivariate-normal region written with the Q function, the standard deviation of a two-regime normal mixture, an expected fine that needs a binomial third moment, and a Monte Carlo tail estimate for a sample variance (at most 100,000 iterations, 2 significant figures). The technology test, which the firm says can be multiple choice or open-ended, mixes C++ and Python output prediction with two easy problems and one harder scheduling problem, sometimes plus a fair-coin z-test or a disk-density transform.
What it tests
- Lasso versus ridge, feature selection and the sources of prediction error
- Gaussian integrals, normal-region probabilities and the Q function
- Moments of binomials and normal mixtures, done fast and right
- A correct Monte Carlo estimate under a time and iteration cap
- Technology track: evaluation order, integer division, aliasing and mutable defaults in C++ and Python
How to prepare
- Part one allows about 30 seconds per question; skip long integrals and return to them.
- Have closed forms ready: the Gaussian integral, binomial moments, and mixture variance via the law of total variance.
- For the Monte Carlo task, vectorise in numpy, fix a seed and check against any closed form you can get.
- Engineers: drill C++ and Python output prediction before the coding half.
Format from compiled accounts of the research test (one 2026 sitting) and the technology test; the firm's pages name the HackerRank test; one developer-role report mentions it.
03Recruiter interview
Phone or video conversation with a recruiter, after the test on the firm's technology and graduate tracks
The firm's pages say this call covers motivation, business knowledge and cultural fit. Candidates report being asked what the firm does, why it and why finance, and how they handle a disagreement in a team. Compiled accounts describe it as an intro on background and interest, sometimes with a short technical warm-up.
What it tests
- A specific answer to why this firm and why trading
- Basic knowledge of what an ETF market maker does
- Motivation and team fit
- Logistics: track, office and start date
How to prepare
- Read the firm's latest annual results and know its businesses: ETPs, fixed income and digital assets.
- Explain in two minutes how your ML or research work could improve pricing or hedging at an ETP market maker.
- Confirm which track (quant research or technology) and which test you are on.
Motivation and fit questions in four other-role candidate reports (three developer, one trader), as this call is shared by every track; the firm's pages list a recruiter interview on motivation and fit.
04Technical interview: probability and brainteasers
About 45-60 min by video with a researcher or trader
Compiled accounts describe fast expected-value and conditional-probability questions plus brainteasers and sequence puzzles, with the hardest variants kept for the final rounds. Research candidates have reported a mix of probability and brainteasers and an open question on how to compute pi. Reported themes include lopsided bets, rolling until a sum is exceeded, Bayesian updates from a trade or a noisy signal, broken-stick geometry and Gaussian integrals.
What it tests
- Expected value of quick games and stopping rules
- Conditional probability and Bayesian updating
- Geometric probability: broken sticks and random points
- Estimating a constant such as pi by a series or by simulation
- Reasoning aloud as you go
How to prepare
- Give an estimate first, then refine; interviewers want to hear the reasoning.
- Practise Bayes updates where the evidence is a trade or the lack of one.
- Have two ways to compute pi ready: a series and a Monte Carlo with its error rate.
Free reported questions for this stage
- Triangle From Breaking the Longer PieceProbabilityMediumSolve
Two quant researcher candidate notes (2019, 2021) cite probability, brainteasers and computing pi; format from compiled accounts of the firm's quant research loop.
05Technical interview: statistics, regression and ML
Technical interview with the research team, part of the team interview the firm says covers technical skills and relevant knowledge
Research candidates have reported regression questions, such as the largest R-squared a two-variable regression can reach when each single-variable regression has R-squared 0.1, and optimisation questions on fitting many parameters, backpropagation and the learning rate. The postings stress statistics, linear algebra, finding signals in large datasets, supervised learning and backtesting, and the first half of the research test (lasso, ridge, feature selection, covariance) previews the topics.
What it tests
- Regression geometry: R-squared with correlated regressors, collinearity and suppression
- Regularisation and shrinkage on noisy, non-stationary market data
- Gradient-based fitting: backprop, learning rates and many-parameter problems
- Time-ordered validation and leakage in backtests
- Signal decay and when a model needs refitting
How to prepare
- Work the two-regressor R-squared bound by hand: it can far exceed 0.2 when the regressors are correlated with each other.
- Explain why a shuffled k-fold split leaks on time series and how walk-forward or purged validation fixes it.
- Link lasso and ridge to priors and to how many independent signals noisy data can support.
Two quant researcher candidate notes (2021 on backpropagation and learning rate, 2025 on regression R-squared); topics also from the research test in compiled accounts and the firm's postings.
06Business knowledge: ETFs and market making
Usually folded into the recruiter and team interviews for research roles; compiled accounts describe a 45-60 min trading-intuition round for trading candidates
Compiled accounts list ETF creation and redemption, leveraged-ETF daily rebalancing, basic options and quoting a two-sided market. A research candidate was asked simply what an ETF is, and the firm's pages say interviews weigh business knowledge on every track.
What it tests
- What an ETF is and how creation and redemption keep it near fair value
- Why a leveraged ETF drifts from its multiple through daily rebalancing
- Quoting a two-sided price and the risk of being picked off
- Basic option deltas and path dependence
How to prepare
- Read a short ETF primer; the firm's own FAQ recommends an ETF handbook to trader applicants.
- Work the arithmetic of a 3x ETF over two volatile days.
- Explain adverse selection plainly: who trades with you when your quote is wrong.
One quant researcher candidate note was asked what an ETF is; the firm's pages say interviews assess business knowledge; round content from compiled accounts.
07Codepair and technical interview (engineering roles)
Live codepair with a senior developer; the graduate developer track uses C++ or Java, research engineer roles are Python-heavy
The firm's graduate software track runs a codepair after the recruiter call to assess problem-solving and coding. Developer candidates report low-level C++ questions on the runtime cost of common operations, STL containers, hashing and sorting, and move semantics for latency. Compiled accounts list a streaming median, inversion counting, 2D prefix-sum reconstruction and an order-matching engine among reported coding problems. Research candidates may also be asked to implement a model or algorithm in Python or C++, which the postings require.
What it tests
- Clean, correct code for data-structure problems such as a streaming median
- Divide-and-conquer counting, such as inversions via merge sort
- Performance awareness: memory layout, containers, copies versus moves
- Turning a research idea into tested code
How to prepare
- Write the obvious correct version first, state its complexity, then optimise.
- Know the cost of Python, numpy and polars operations as well as C++ ones.
- Talk through tests and edge cases before you run anything.
Free reported questions for this stage
- Counting Inversions via Modified Merge SortCoding & AlgorithmsMediumSolve
The firm's graduate developer page lists a codepair with a senior developer; three developer-role candidate reports describe C++ questions but are not ML-role reports, so they are not counted.
08Final round: research deep dive and team lead
Final interview with a team lead or desk head plus team members; in person, with travel booked by the firm when needed
The firm's pages describe a final interview with a team lead on technology tracks and with the head of the desk and team members on trading tracks. Research candidates report being asked to explain their research and walk through their last project. Compiled accounts put the hardest probability and puzzle variants here, and a 2024 graduate developer report had brainteasers in the final.
What it tests
- Explaining your research: question, method, result and what failed
- Defending choices under follow-up questions
- Harder versions of earlier probability problems
- Team compatibility and the firm's values
How to prepare
- Prepare one project as a ten-minute talk and as a two-minute version.
- Expect questions on how your result would hold up out of sample and in live use.
- Keep a few hard Gaussian and conditional-probability problems fresh for this round.
Two quant researcher candidate notes (2022) describe explaining research and a last project; the firm's hiring pages list a final interview with a team lead or desk head.
Questions and themes reported most often
- Why this firm, what it does and why finance (3 reports, recruiter interview)
- Explain your research or walk through your last project (2 reports, final round: research deep dive and team lead)
- Largest R-squared of a two-regressor model when each alone gives 0.1 (1 reports, technical interview: statistics, regression and ml)
- Fitting many parameters: backpropagation and the learning rate (1 reports, technical interview: statistics, regression and ml)
- How would you compute pi (1 reports, technical interview: probability and brainteasers)
- What an ETF is and why leveraged ETFs drift (1 reports, business knowledge: etfs and market making)
- Two-part research test: stats and ML multiple choice, then probability with a Monte Carlo task (online assessment)
- Lasso versus ridge and feature selection (online assessment)
What Flow Traders asks about
- Probability & expected value28%
- Statistics, regression & ML27%
- Coding (test + codepair)20%
- Research deep dive & behavioral15%
- ETF & market basics10%
Flow Traders interview prep checklist
- Finished a 30-question, 15-minute statistics and ML multiple-choice drill
- Solved six short probability questions plus a Monte Carlo task in 30 minutes
- Can derive the two-regressor R-squared bound and contrast lasso with ridge
- Can explain time-ordered validation and leakage for a trading signal
- Can explain an ETF, creation and redemption, and leveraged-ETF drift
- Prepared a ten-minute research deep dive with out-of-sample evidence
- Engineers: predicted the output of 20 tricky C++ and Python snippets and did a timed codepair
Tips
- Ask which test you will get: research applicants report a statistics, ML and probability test, technology applicants a language-quirk and coding test.
- Drill timed ML and statistics multiple choice: lasso and ridge, feature selection, bias and variance, covariance matrices, Brownian motion.
- Make Gaussian integrals, the Q function, and binomial and mixture moments automatic.
- Write a short numpy Monte Carlo with a fixed seed and an error estimate until it takes under five minutes.
- Work regression geometry problems such as R-squared bounds with correlated regressors.
- Prepare a research talk that stresses out-of-sample validation and what happens in live use.
- Learn ETF basics: creation and redemption, leveraged-ETF rebalancing and a two-sided quote.
- Skip mental-math sprints and sequence drills unless you also apply to the trader track.
- Engineers: practise C++ and Python output prediction and a live codepair with tests.
See the rest of the Flow Traders loop
The later stages, the questions reported for them and how to prepare are in Pro.
Unlock with ProPractice the Flow Traders loop
Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.
Flow Traders interview FAQ
How many interview rounds does Flow Traders have for ML roles?
Candidates report 8 stages: Application and resume screen, Online assessment, Recruiter interview, Technical interview: probability and brainteasers, Technical interview: statistics, regression and ML, Business knowledge: ETFs and market making, Codepair and technical interview (engineering roles) and Final round: research deep dive and team lead. Apply with a CV and cover letter, then an online test: research applicants report two take-homes back to back (a 15-minute statistics and ML multiple-choice section, then 30 minutes of probability with one Python Monte Carlo task), while technology applicants take a test of language-quirk multiple choice plus three coding problems. Next comes a recruiter interview on motivation, business knowledge and fit, then technical interviews with the team: probability and brainteasers, statistics and regression depth, and for engineers a codepair with a senior developer. The final round is with a team lead or desk head and centres on your past research and fit; team allocation is decided during the interviews, and the firm books travel when finals are in person. The numerical mental-math and sequences test and the senior-trader case study are trader-track only, and there is no market-making game. The firm's pages put the test before the recruiter call, while compiled accounts list a recruiter screen first.
How long does the Flow Traders ML interview process take?
Reported timelines run Varies; neither the firm nor compiled accounts give a typical length.
What does Flow Traders ask in ML interviews?
By reported question mix: Probability & expected value 28%; Statistics, regression & ML 27%; Coding (test + codepair) 20%; Research deep dive & behavioral 15%; ETF & market basics 10%. The most repeated themes: Why this firm, what it does and why finance; Explain your research or walk through your last project; Largest R-squared of a two-regressor model when each alone gives 0.1.
How should I prepare for the Flow Traders ML interview?
Ask which test you will get: research applicants report a statistics, ML and probability test, technology applicants a language-quirk and coding test. Drill timed ML and statistics multiple choice: lasso and ridge, feature selection, bias and variance, covariance matrices, Brownian motion. Make Gaussian integrals, the Q function, and binomial and mixture moments automatic.
Are there free Flow Traders interview questions?
Yes. LeetML has 6 questions reported from Flow Traders interviews, and 2 of them are free to solve in the browser; they are listed by stage above.
Related interview guides
- OptiverFor ML, AI and research roles: a long timed assessment battery, a live betting game with a trader, a probability whiteboard, a P&L-graded take-home and a blunt fit final.
- IMC TradingFor ML, AI and research roles at a global options market maker: a track-split timed test, probability and ML-fundamentals rounds, and a fast superday.
- Virtu FinancialFor ML, AI and research roles: a coding test gate, quick brain teasers, probability and regression phone rounds, then a logic-test onsite and a modeling final.
- Susquehanna (SIG)For ML, AI and research roles: a timed probability test, regression-heavy technical calls, a trade-data take-home and a long onsite.
- Jane StreetFor ML, AI and research roles: multi-part coding judged on clarity, a hands-on ML round, and probability and statistics on research-leaning loops.
- Akuna CapitalFor ML, AI and research roles: a proctored Python coding test, a camera-on math test, probability interviews, then a final round on statistics, ML and your research.
Official sources
- Firm site: careers and FAQ (official)
- Firm site: graduate programmes and hiring steps (official)
- Firm site: technology hiring process (official)
- Firm site: trading and quant research hiring process (official)
- Job posting: PhD Graduate Quantitative Researcher, Amsterdam (official)
- Job posting: Senior Research Engineer, London (official)