Quant Researcher Interview Questions — and how to actually answer them
Quant researcher interviews at firms like Jump Trading, Citadel, Jane Street, and Two Sigma follow a predictable structure — but most candidates prepare for the wrong things. The goal is not to memorize solutions. It is to demonstrate how you think when you are stuck, uncertain, or being deliberately pressured.
Below is a breakdown of the six main question categories, with concrete examples and guidance on what the interviewer is actually looking for in each.
01
Probability & Brain Teasers
This is the category most candidates are aware of, and the one most consistently mishandled. The questions are not puzzles to be solved silently — they are conversations. Interviewers want to see your reasoning in real time, including the wrong turns.
Example questions
- You flip a fair coin until you see two heads in a row. What is the expected number of flips?
- Two players take turns rolling a fair die. The first to roll a 6 wins. Player A goes first. What is the probability Player A wins?
- You have 100 light bulbs in a row. You walk by every 2nd, then every 3rd, toggling each. How many are on after all passes?
- A fair die is rolled repeatedly. What is the expected number of rolls until you see all six values?
How to approach these: Set up the state space explicitly before solving. For the coin-flip example, define states (no heads yet, one head, done) and write the recurrence before computing. Say what you are doing. For symmetry arguments, name the symmetry. Interviewers penalize candidates who jump to a numerical answer without explaining why it is correct — and reward those who find elegant structure even when they cannot complete the arithmetic cleanly.
One common mistake: candidates confuse E[X] with 1/p when problems have dependent structure. Always check whether successive trials are truly independent before applying the geometric distribution shortcut.
02
Statistics & Estimation
Quant researcher roles require applied statistical intuition, not just theorem-recitation. Interviews test whether you understand the assumptions behind common methods and what happens when those assumptions break.
Example questions
- Walk me through how you would test whether a trading strategy has real alpha versus just in-sample noise.
- You fit a linear regression to financial data. The residuals show strong autocorrelation. What are the consequences, and what would you do?
- How would you estimate the volatility of a stock that has only 30 days of history? What are the limitations of your estimate?
- What is the difference between p-hacking and legitimate multiple testing, and how do you adjust for it?
How to approach these: The alpha-testing question is a filter for candidates who have worked with real data versus those who have only studied textbooks. Good answers mention out-of-sample testing, overfitting risk, the multiple comparisons problem, and the Sharpe ratio's distributional assumptions under non-normality. Weaker answers go straight to a t-test without discussing any of these concerns.
For the regression question, know that autocorrelated residuals do not bias OLS coefficient estimates but do invalidate standard errors — which matters a great deal if you are building confidence intervals on signal strength. Knowing what breaks and what does not is the answer.
03
Coding (Python / C++)
Coding rounds for quant researcher roles differ from SWE interviews. The problems are shorter and more numerically focused. Interviewers care about correctness, clarity of reasoning, and awareness of numerical edge cases. A clean O(n log n) solution with correct edge-case handling beats a clever but unreadable O(n) solution.
Example questions
- Implement a function that computes a rolling mean and rolling standard deviation over a window of size k, efficiently.
- Given a list of trade timestamps and prices, compute the time-weighted average price (TWAP).
- Write a function to detect whether a time series has a statistically significant trend (without using a library).
- Implement a simple order book that supports insert, cancel, and best-bid/best-ask queries.
How to approach these: The rolling statistics question is a warm-up for checking whether you know about numerical stability. A naive two-pass approach is fine to state first; then the interviewer will push you toward Welford's online algorithm or a deque-based approach. Know why naive variance computation loses precision with large means and small variance.
For C++ roles especially: understand move semantics, cache locality basics, and why std::vector beats std::list for most trading applications. You will not be expected to write production-grade lock-free code in 30 minutes, but you should know why it matters.
04
Mathematics: Stochastic Calculus & Linear Algebra
For researcher roles at systematic firms, mathematical depth is a genuine filter. Stochastic calculus questions assess whether you have real intuition for continuous-time models, not just formula-memorization. Linear algebra questions test the foundations of factor models, PCA, and regression — methods you will use constantly.
Example questions
- State Itô's lemma. Apply it to derive the SDE for ln(S) when S follows geometric Brownian motion.
- What is the quadratic variation of Brownian motion over [0,T]? Why does it matter for option pricing?
- You have a 1000×50 matrix of returns. Explain how you would use PCA to construct a factor model. What does each component represent?
- Why does ridge regression shrink coefficients toward zero? In what scenarios is it better than OLS, and when does it fail?
How to approach these: On stochastic calculus questions, start from the definition and derive rather than recite. Interviewers at systematic firms have heard the formula a thousand times — what they want is to see you derive d(ln S) from Itô's lemma step by step, including the correction term, and explain why that correction term is non-zero.
On PCA: be specific about what "explaining variance" means in the context of returns. The first principal component of an equity return matrix is usually the market factor. Be ready to discuss the difference between PCA of a covariance matrix vs a correlation matrix, and why the choice matters for factor construction.
05
Market & PnL Reasoning
These questions do not require knowledge of proprietary strategies. They test whether you can think in terms of a trading firm's economics: how positions are sized, where risk comes from, what adverse selection looks like, and how to reason about signal decay. They are common at HFT and prop shops even in researcher (not trader) roles.
Example questions
- You discover a signal with a Sharpe ratio of 1.5 on 5 years of data. How do you decide whether to trade it live?
- What is adverse selection in the context of market-making? How would you detect it in your fill data?
- A strategy has a positive expected value per trade but goes through a 30-day drawdown. How do you distinguish bad luck from a regime change?
- How would you think about position sizing if your signal has varying confidence across different instruments?
How to approach these: The Sharpe question is a structured critical-thinking exercise. A strong answer discusses the statistical significance of a Sharpe estimate (hint: 5 years of daily data is about 1,250 observations — the standard error on Sharpe is roughly 1/√n), the risk of overfitting to the backtest period, and what out-of-sample validation would look like. It does not treat 1.5 as inherently impressive.
For the drawdown question, good candidates distinguish between a drawdown that is within the expected range given the strategy's historical distribution and one that flags a structural break. This requires knowing the strategy's maximum drawdown distribution — not just its Sharpe — before going live.
06
Behavioral & Motivation
Systematic trading firms hire for cultural fit as seriously as technical ability. The behavioral round is not a formality — a strong technical candidate can be rejected on behavioral grounds. These firms operate with small teams, large autonomy, and high stakes. They want people who are intellectually honest, low-ego, and rigorous about their own limitations.
Example questions
- Describe a project where you were certain you were right but turned out to be wrong. What happened?
- Tell me about the most technically complex thing you have built or analyzed. Walk me through it.
- How do you decide when to stop investigating a research question?
- What do you do when you disagree with a senior researcher's approach to a problem?
How to approach these: The "wrong about something" question is a test of intellectual honesty. Interviewers have heard the answer where the candidate was "wrong" about something minor that turned out fine anyway. They want a real example: something where you had conviction, pursued it, and discovered you were off — and where you can explain what the actual error was at a technical level.
On "when to stop a research question": this is a filter for obsessive vs. disciplined researchers. The answer interviewers want is something like: when the expected information gain from further investigation is lower than the cost of the next experiment, given your current uncertainty. Be concrete about how you make that judgment call.
Firm Differences
How the process differs by firm
The six categories above are universal. The emphasis varies:
Jump Trading
Probability and brain teasers are extremely heavy in early rounds. Math depth is expected at the researcher level — Itô, SDEs, and their applications are fair game. Coding is typically Python but increasingly C++ for lower-latency roles.
Jane Street
Market reasoning and trading intuition are central alongside math. Expect questions about making markets, options pricing intuition, and probability under time pressure. The culture values clear verbal reasoning.
Citadel / Citadel Securities
Statistics and signal research methodology are prominent in researcher tracks. Expect questions about backtesting methodology, factor model construction, and applied machine learning in finance.
Two Sigma
Strong emphasis on coding quality and system design alongside research. Candidates often encounter problems that blend statistics, data analysis, and implementation in the same round.
These characterizations reflect publicly known information and general industry patterns. Firm-specific hiring processes change over time, and any session with Adrien covers only what is publicly known or non-proprietary.
If you are looking for a structured week-by-week approach, the realistic 8-week quant trading interview prep plan covers how to sequence these categories and what to prioritise depending on your timeline. If you are still deciding between the researcher and trader track, see the quant trader vs quant researcher comparison for a breakdown of how the roles differ and how to choose.
Practice with someone who has been inside
Book a Mock Interview or Deep Dive
Reading through question categories is useful preparation. Working through them live, under realistic pressure, with direct feedback on where your reasoning breaks down, is a different order of magnitude. Adrien — former quantitative researcher at Jump Trading, École Polytechnique, Stanford MS in Computational Mathematics (ICME), 3× IMO medalist for France — has been on both sides of these interviews.
Deep Dive
POPULAR60-minute video call
- 60-minute 1:1 technical session
- Work through real problem types from the categories above
- Resume and positioning review
- Firm-specific strategy
Mock Interview
60 min + written feedback
- 60-minute full mock interview simulation
- Questions across all six categories
- Realistic pressure and pacing
- ~1 page written feedback within 48h
NDA-bound from Jump Trading — cannot discuss specific strategies, signals, or proprietary infrastructure. Sessions available in English and French.