Question 1

How is a quant interview different from a software engineering interview?

Accepted Answer

Quant interviews weight probability, statistics and mental mathematics far more heavily, and often include market-making or betting games that test decision-making under uncertainty. Quant developer roles add serious C++ and low-latency systems depth. The three tracks differ: quantitative researcher is mathematics and statistics heavy, quant developer is systems and C++ heavy, and trader is games, mental maths and risk judgement. Prepare for the one you applied to — the preparation is not interchangeable.

Question 2

What probability topics come up most?

Accepted Answer

Conditional probability and Bayes, expectation and linearity of expectation, variance, common distributions (binomial, Poisson, normal, exponential), random walks, Markov chains and optimal stopping. Linearity of expectation is the single highest-leverage tool — it works even when events are dependent, which is what makes hard-looking problems collapse. The classic worked example is the expected number of coin flips to see a given pattern.

Question 3

Which books actually help for quant preparation?

Accepted Answer

Xinfeng Zhou's "A Practical Guide to Quantitative Finance Interviews" — widely called the green book — is the standard. "Heard on the Street" covers brainteasers and pricing; "Fifty Challenging Problems in Probability" builds the core intuition. Work the problems before reading the solutions. Reading the green book cover to cover teaches you very little; solving 100 of its problems under time pressure is what changes performance.

Question 4

How do I prepare for mental maths tests?

Accepted Answer

Practise timed arithmetic daily — typically 60 to 80 questions in eight minutes, mixing multiplication, percentages, fractions and decimals. Learn the shortcuts: squares to 30, fraction-to-decimal conversions, and percentage manipulation. Firms use this as a hard filter, and it is purely trainable. Two weeks of ten minutes a day usually moves someone from failing to comfortably passing — it rewards practice more than talent.

Question 5

What is a market-making game and how should I play it?

Accepted Answer

You are asked to quote a two-sided price — a bid and an ask — on some uncertain quantity, and the interviewer trades against you. You must update on the information their trades reveal and manage your position. Start with a wide spread because you are uninformed, and narrow it as you learn. The critical skill is inference: if they keep buying at your ask, your price is too low — adjust rather than repeating it. Being wrong is fine; failing to update on evidence is not.

Question 6

What C++ depth does a quant developer intern need?

Accepted Answer

Genuine depth: memory management and RAII, move semantics, when to use each smart pointer, virtual dispatch and its cost, templates, cache behaviour, and what actually happens at each level of optimisation. The distinguishing questions are about cost, not syntax — why a virtual call is not free, why a vector outperforms a list even for insertions, why false sharing destroys multi-threaded performance. Answers grounded in cache lines and branch prediction score highest.

Question 7

What should I know about low-latency systems for an HFT internship?

Accepted Answer

Cache hierarchy and locality, avoiding allocation on the hot path, lock-free structures, kernel bypass networking at a conceptual level, and how you would measure latency including percentile tails rather than averages. Nobody expects an intern to have written a matching engine. They do expect you to know that the mean latency is the wrong metric and that p99 is where the money is lost, and to reason about why a cache miss costs hundreds of cycles.

Question 8

How do I approach a brainteaser I have never seen?

Accepted Answer

Restate it, check for a symmetry or invariant, try the smallest case, then generalise. Narrate your reasoning — the process is what is being graded, and interviewers routinely hint if you are audibly working. Many brainteasers reduce to linearity of expectation, a symmetry argument, or working backwards from the end state. Guessing an answer without reasoning scores zero even when correct, because it demonstrates nothing.

Question 9

What Python do algorithmic trading internships expect?

Accepted Answer

NumPy and pandas fluency, time-series handling including resampling and alignment, a clean backtest that avoids lookahead bias, and honest performance statistics — Sharpe, drawdown, turnover. Lookahead bias is the standard trap and the standard interview question. If your backtest uses a day's closing price to make that day's decision, the strategy is fictional. Being able to explain how you prevented it is the point.

Question 10

What degree backgrounds do quant firms hire from?

Accepted Answer

Mathematics, statistics, physics, computer science and engineering are all common; physics and mathematics are strongly represented in research roles. What matters is demonstrable mathematical maturity, not the specific degree name. These processes are academically selective and competition is intense. If your mathematics background is lighter, quant developer roles are usually the more realistic entry point than quantitative research.

Question 11

How should I answer a question about risk in a trading interview?

Accepted Answer

Talk about position sizing relative to bankroll, expected value versus variance, and the scenario that would ruin you rather than the one that pays. Show that you think about the downside first. A candidate who maximises expected value while ignoring ruin risk fails this round. Referencing the Kelly criterion, and then noting that practitioners bet a fraction of Kelly because of estimation error, is a strong, grounded answer.

Question 12

What statistics questions come up for a quantitative researcher intern?

Accepted Answer

Hypothesis testing and p-values, confidence intervals, linear regression and its assumptions, overfitting and cross-validation, multiple testing, and stationarity in time series. Multiple testing is the one most connected to the actual job: if you test 500 signals at the 5 percent level, roughly 25 will look significant by chance. Being able to explain that, and what you would do about it, is directly relevant to research work.

Quant, Trading and Financial Engineering Interviews Interview Questions and Answers

How to use this set