Quant Researcher Interview Questions & Prep (2026 Guide)
Quantitative Finance
Exponent Team • Published Quant researcher interviews go deeper on statistics than any other quant role, and the research case study is where most offers get decided. A timed probability and statistics screen comes first at most firms, before any researcher speaks with you. Past that screen the process diverges, and Citadel, Two Sigma, D.E. Shaw, G-Research, WorldQuant, AQR, and 4 other firms each test a different mix.
This guide collects the real questions each one asks and shows what changed in the 2026 rubric. You'll get a sequence for open research prompts and 8-week and 2-week prep plans.
What is a quant researcher interview?
A quant researcher interview tests how you design and validate a predictive signal. Interviewers score the validation work as closely as the model itself. The buy-side research loop covers probability and statistics, machine learning and modeling, Python coding, and mathematics. It closes with at least one research case study where you frame a hypothesis, build something, and defend it. The coding is Python and data-analysis work. Quant developer loops test low-latency systems, so confirm which track you're in before you prepare.
The exact shape of the quant researcher interview varies by firm. A common format opens with a timed online assessment or quant quiz, then 1 to 2 technical phone screens on probability and statistics. An onsite follows, sometimes a full-day Superday covering coding, statistics, ML, and a research discussion. A behavioral or portfolio-manager round closes the loop. The probability and statistics gate usually comes first, before any case work, so firms can filter on fundamentals before a human spends time with you.
Level changes which rounds decide the outcome. New-grad and intern processes turn on the online assessment and the fundamentals rounds, where speed and accuracy carry most of the signal. Senior and PhD-track processes turn on your own work, with a detailed review of signals you've built and how you validated them.
| Round | What it tests | Typical format |
|---|---|---|
| Online assessment / quant quiz | Speed and accuracy on probability, statistics, and coding fundamentals | Timed, often multiple choice or auto-graded code |
| Technical phone screen (1-2) | Probability and statistics reasoning, follow-up depth | 45 to 60 min with a researcher |
| Onsite / Superday | Coding, statistics, ML modeling, and a research case study | 4 to 7 back-to-back rounds |
| Behavioral / PM / MD | Fit, intellectual honesty, "why this firm" | Conversational |
The round count varies most at the onsite stage, where a full Superday can run 4 to 7 interviews back to back and a leaner process stops at 3.
Quant researcher interviews in 2026
Quant researcher interviews changed in 2026 by adding research hygiene to the scoring rubric: interviewers now test how you validate a result as closely as how you build the model. Machine learning and alternative data cover more of the loop, timed quizzes screen fundamentals before any human round, and LLM tooling moved the bar toward judgment work a model can't supply. The questions themselves changed less than the scoring did.
| Trend | What interviewers now score | What to prepare |
|---|---|---|
| Machine learning and alternative data | Model selection, feature engineering, and the validation of noisy signals | A signal you built from messy data, and why you trust it |
| Research hygiene | Whether you can avoid a false positive | Walk-forward validation, purging and embargo, false-discovery rate |
| Timed quant quizzes | Speed and accuracy on fundamentals, before any human round | Probability and programming under a clock, trained as its own track |
| LLM-assisted workflows | The judgment and validation a model can't supply | Why you chose a validation scheme, and what it would miss |
Two Sigma integrates ML and alternative data directly into its research. Expect questions on model selection, feature engineering, and how you'd validate a signal built from a noisy source. Most hedge funds now use alternative data in some form, so the judgment questions matter more than the tooling.
D.E. Shaw presses on false-discovery rate and on t-stats corrupted by autocorrelation, and Two Sigma tests your validation scheme directly. Some of that shift traces to large language models, which now handle literature review, boilerplate backtest code, and early data exploration. Expect follow-ups on why you chose a validation scheme and what it would miss.
G-Research screens with a multiple-choice quiz, and WorldQuant's online assessment mixes probability and programming. Both sit ahead of the technical rounds, so speed and accuracy on fundamentals decide whether you reach the rounds that test judgment.
Quant researcher interview questions
Quant researcher interview questions cluster into 6 areas: probability and statistics, machine learning and modeling, Python coding, mathematics, research case studies, and behavioral. They overlap with broader quant finance questions, and they go deeper on statistics and on proving a result holds up.
Probability and statistics
Probability and statistics questions decide more quant researcher screens than any other question type. Expect closed-form probability, estimator theory, and follow-ups that escalate until you reach the limit of what you know.
Expect questions like:
- A geometric random variable question at D.E. Shaw: the expected number of die rolls until the first 6.
- Bayes' theorem applied to a practical scenario in Citadel's intern rounds, followed by an open brainteaser.
- OLS assumptions, the derivation of beta, and when the estimator is BLUE, which AQR asks in close to every round.
- Conditional expectation and hypothesis testing with escalating follow-ups at Point72 and Cubist.
How D.E. Shaw scores the die-roll question:
- Core knowledge is tested: Recognizing a geometric random variable and computing its expectation
- How long the answer should take: Roughly 2 to 3 minutes. It's a warm-up question.
- What the interviewer is looking for: A clean setup (each roll is independent, success probability 1/6) and the result, E[X] = 1/p = 6
- Good vs. great answers: A good answer gives 6. A great one explains why E[X] = 1/p and extends it, for example to the expected rolls to see all 6 faces (the coupon-collector follow-up).
Machine learning and modeling
ML rounds test design judgment more than algorithm recall, and the modeling fundamentals overlap with what our ML engineer course covers. You'll pick a model, justify the choice against a simpler baseline, and say how you'd know the result generalizes.
Expect questions like:
- Pseudocode for k-means at Point72 and Cubist, with a follow-up on its convergence guarantee.
- A model design question at Two Sigma: predict Manhattan rent prices.
- Feature engineering, model selection, regularization, and tree-based vs. linear models, all of which Citadel and Cubist cover.
- Cross-validation strategy at both firms, alongside spotting overfitting and leakage.
Coding (Python)
Quant researcher coding rounds test data-analysis fluency under a clock. Cubist runs practical data-analysis coding, and Two Sigma asks for production-quality code in a timed setting.
Expect questions like:
- An efficient linear regression implementation in Two Sigma's online assessment.
- Rolling statistics in pandas and numpy, written to run over a full series.
- A simple estimator built from scratch, with its complexity explained.
Mathematics
Math rounds stay close to what the research job uses: linear algebra, calculus, stochastic processes, and time series. G-Research and Citadel both cover this ground, and several of the exact questions in circulation have been asked for years.
Expect questions like:
- An OLS proof at Two Sigma, framed as an optimization with Lagrange multipliers.
- The product of 2 very large matrices at Citadel, computed efficiently by using their structure.
- Markov chain questions of the kind collected in the "Green Book."
Research case studies
The case study separates quant researcher interviews from every other quant role. You'll frame a hypothesis, build something, and defend the design against an interviewer looking for the hole in it.
Expect prompts like:
- Signal construction and backtest design at Two Sigma, including validation that blocks lookahead and leakage.
- Open-ended framing, such as how you'd make money using social-media data.
- A review of your own signals in Millennium's senior and PhD-track rounds: how you built them, what data you used, and how you ran the backtest.
Behavioral
Behavioral rounds at quant firms test intellectual honesty more than culture fit. D.E. Shaw asks "why this firm" and "why finance" with follow-ups until the answer gets specific, and a generic answer about wanting hard challenges ends the conversation early.
Prepare 2 answers before the interview: one on why finance over academia or tech, and one on a result you had to abandon. Cubist's portfolio-manager round asks how you handle being wrong, so bring a research decision you got wrong and can explain.
Quant researcher interviews by company
Quant researcher interviews differ by firm in 2 ways that change how you prepare: what the first filter is, and how much of the process reviews research you've already done. Citadel, Two Sigma, D.E. Shaw, G-Research, WorldQuant, Point72/Cubist, AQR, Squarepoint, Millennium, and Jane Street hire the most quant researchers, and the interview process is different for each.
| Firm | First filter | Signature round |
|---|---|---|
| Citadel | Online assessment | 4 to 6 round Superday |
| Two Sigma | Technical phone screens | Up to 7 rounds across technical, hiring-manager, and MD |
| D.E. Shaw | 45-minute screen on your research | Probability-heavy loop of 5 to 7 interviews |
| G-Research | 90-minute multiple-choice quiz | 4 one-hour technical interviews |
| WorldQuant | Online assessment (probability and programming) | Interview in a technical focus you pick |
| Point72 / Cubist | Online assessment | Research-judgment round, sometimes a take-home |
| AQR | Coding online assessment | PnL-calculator take-home |
| Squarepoint | Timed coding challenge | Probability question, then a research discussion |
| Millennium | Pod-dependent, often a resume screen | 24-hour Jupyter notebook |
| Jane Street | No published format | Assumptions that change mid-question |
Cubist, Point72's systematic arm, is distinct from the firm's discretionary long/short teams. Its interview process covers probability, statistics, ML, and Python coding, plus a research-judgment overlay and sometimes a take-home modeling exercise. Squarepoint opens with a timed coding challenge, then a probability question that scores your reasoning above a memorized formula, then a research discussion. It's developer-heavy, so expect solid coding alongside the statistics.
Jane Street tests broad intellectual curiosity and changes the assumptions mid-question to see how you update. Jane Street also discourages candidates from discussing specifics, so expect questions you won't find in any public list.
Citadel
Citadel pairs a probability and statistics focus with ML and clean coding across a 4 to 6 round Superday, and its research seats are in-office. Expect a statistics question on evaluating a trading strategy. The online assessment comes first, then a technical round with a medium-to-hard coding challenge.
Two Sigma
Two Sigma's loop is broad by design: probability, statistics, ML, SQL, Python, and case studies in one process. Expect up to 7 rounds: 3 technical, 3 with hiring managers, and one with an MD. The breadth is the challenge here, so prepare across statistics and modeling, and don't specialize.
D.E. Shaw
D.E. Shaw runs the most academically rigorous process, opening with a 45-minute screen on your research, why finance, and why the firm, then 5 to 7 interviews over 4 to 8 weeks. Probability and statistics take the largest share of that loop, with Python data analysis and research design filling most of the rest. Expect follow-ups that test self-skepticism, including false-discovery rate and autocorrelation.
G-Research
G-Research's signature filter is a 90-minute multiple-choice quiz, with a general quant version or an ML-specific one, followed by 4 one-hour technical interviews and leadership rounds. G-Research recommends Kaggle as preparation.
WorldQuant
WorldQuant centers on alphas, the predictive signals built on its BRAIN platform, and moves fast, often in about 2 weeks. The process starts with an online assessment mixing probability and programming, then an interview where candidates pick a technical focus (math, programming, statistics, data science, or financial math).
AQR
AQR's research loop centers on factor investing (value, momentum, carry, defensive) and on OLS, which appears in almost every round. The path moves from a coding online assessment to a CS-fundamentals phone screen, a PnL-calculator take-home, and an onsite. Expect a prompt on how you'd handle a trading model with 25 factors compared with one with 100, and what changes if the 100-factor model performed 25% better.
Millennium
Millennium's independent pods each run their own hiring, so the format varies more here than at other firms. Some pods run 4 rounds: a 24-hour Jupyter notebook, a technical interview, a coding round, and a conversation with senior researchers. Others open with a 45-minute resume-based screen covering Python and probability.
The ALPHA framework for the research case study
An open research prompt in a quant researcher interview runs through 5 stages: ask, load, prototype, harden, apply. Work them in order and you'll cover what the case round tests, including the validation step that carries the most scoring weight in 2026. ALPHA is the shorthand for that sequence.
| Step | What you do | What it tests |
|---|---|---|
| Ask | Frame the hypothesis and define the success metric before touching data | Framing and judgment |
| Load | Get and clean the data, engineer features, and respect point-in-time availability | Data judgment |
| Prototype | Build a first model or signal, simple before fancy | Modeling fluency |
| Harden | Validate out of sample, and guard against overfitting, lookahead, leakage, and false discovery | Research hygiene (the 2026 focus) |
| Apply | Turn the signal into something tradeable, monitor it, and defend your choices | Production sense |
ALPHA describes the case round, and it isn't the first thing you face. Most firms screen fundamentals before you reach a case at all. Use ALPHA as a checklist for your own reasoning, and expect each firm to compress or reorder the steps.
How to prepare for a quant researcher interview
Quant researcher prep covers 3 tracks: probability and statistics for the timed screen, modeling and validation for the ML and case rounds, and your own past research for the defense round. Work all three, since each one is scored in a different round.
Where to focus:
- Work probability and statistics first and daily. Use Zhou's "Green Book" for breadth, and Casella & Berger if you're on a PhD-track process.
- Practice OLS end to end: the assumptions, the derivation of beta, and when the estimator is BLUE. AQR asks it in close to every round.
- Work time series with Box-Jenkins, since stationarity and autocorrelation questions turn up throughout the statistics rounds.
- Prepare 2 or 3 of your own research projects, about 30 minutes each, with the alternatives you considered, the weaknesses you know about, and the extensions you'd try.
- Practice defending a validation scheme out loud, since interviewers press on how you'd catch a false positive in your own work.
- Use Kaggle for modeling practice, which G-Research recommends to its candidates.
- Add timed pattern-based coding practice at medium-to-hard difficulty, since several firms open with a coding assessment.
- Read Ernie Chan's "Quantitative Trading" if you're coming from academia and need the market context.
With 8 weeks, spend the first 4 on probability, statistics, and OLS, then the next 4 on your own projects and timed coding. With 2 weeks, work the Green Book's probability and statistics sections, practice OLS end to end, and walk through 2 of your own projects out loud with their weaknesses ready.
Key concepts every quant researcher should know
Quant researcher interviews assume you know alpha, OLS, overfitting, lookahead bias, walk-forward validation, and factor investing without hesitation. Each concept below appears either in the question buckets above or in the research case study, and the case round tests several at once.
| Concept | What it is and when it applies |
|---|---|
| Alpha | A predictive signal expected to earn returns beyond a benchmark. The object of nearly every research case. |
| OLS / linear regression | The workhorse estimator for alpha research. Know its assumptions, how to derive beta, and when it's BLUE. |
| Overfitting | Fitting noise and calling it signal. The central failure mode in alpha research. |
| Lookahead bias / leakage | Using information not available at prediction time. It invalidates a backtest. |
| Walk-forward validation (purging / embargo) | Time-series cross-validation that prevents leakage across the train/test boundary. |
| Factor investing | Building portfolios around systematic return drivers: value, momentum, carry, defensive. |
| Sharpe ratio | Risk-adjusted return, excess return divided by volatility. |
| Multiple testing / false discovery rate | Controlling for spurious "significant" results when you test many hypotheses. |
| Stationarity / autocorrelation | Time-series properties that break i.i.d. assumptions and inflate t-stats. |
| Regularization (L1 / L2) | Penalizing model complexity to improve out-of-sample generalization. |
| Backtest | A historical simulation of a strategy's performance. |
| Cross-sectional vs. time-series signal | Ranking assets at one point in time vs. predicting one asset over time. |
Quant researcher interview FAQs
What questions are asked in a quant researcher interview?
Common quant researcher interview questions include expected-value and Bayes probability, OLS assumptions and the derivation of beta, k-means and model design, and efficient Python implementations. D.E. Shaw asks for the expected number of die rolls until the first 6, and Two Sigma asks you to design a model predicting Manhattan rent prices. Open research prompts appear at every firm, such as designing a backtest that blocks lookahead bias.
How hard is the quant researcher interview?
The quant researcher interview is among the hardest in finance, combining graduate-level statistics, machine learning, and clean coding under time pressure. You'll face an automated filter first at most firms, whether that's G-Research's 90-minute quiz or D.E. Shaw's probability-heavy screen.
What's the difference between quant researcher, quant trader, and quant developer interviews?
A quant researcher interview centers on statistical modeling and research case studies such as signal construction, backtesting, and overfitting guards. A quant trader interview centers on timed mental math and live market-making exercises, and a quant developer interview centers on C++ and low-latency systems. The same firm can run all three, so confirm your track before you prepare.
How long does the quant researcher hiring process take?
The quant researcher hiring process usually runs 4 to 10 weeks, from online assessment to onsite. WorldQuant is the fastest of the major employers and can close in about 2 weeks, while Citadel, Point72, and Cubist commonly take 6 weeks or more. Ask your recruiter for the expected timeline, since a pod or desk can compress it.
How do you prepare for a quant researcher interview?
Preparing for a quant researcher interview starts with probability and statistics, using Zhou's Green Book and Casella & Berger. Then prepare 2 or 3 of your own research projects in enough detail to answer follow-ups on your data choices and your validation. Timed pattern-based coding practice and Kaggle modeling practice fill the final weeks.
What is the G-Research quant quiz?
The G-Research quant quiz is a 90-minute multiple-choice assessment of 10 questions with 5 options each, taken before any human round. You choose between a general quant version and an ML-specific version. Pass it and you'll sit 4 one-hour technical interviews plus leadership rounds.
Which firms hire quant researchers?
Firms that hire quant researchers include Citadel, Two Sigma, D.E. Shaw, G-Research, WorldQuant, Point72/Cubist, AQR, Squarepoint, Millennium, and Jane Street. Each runs a different process, so confirm both the firm and the track before you prepare.
Your Aced membership awaits.
Aced is the fastest-growing tech interview prep platform. Get free interview guides, insider tips, and courses.
Create your free account