How Quant Interviews Are Structured

Quant trading (market makers & prop firms)

The emphasis is speed and risk intuition under pressure. Expect timed mental-maths screens, number-sequence pattern tests, fast probability and expected-value questions, and a live market-making game where you quote two-sided markets and manage inventory under limits.

• ▸Mental-maths screens: e.g. Optiver's well-known '80 questions in 8 minutes', plus sequence tests.
• ▸Trading games: make a market, update on new information, manage a position — firms such as IMC and Jane Street lean on these.
• ▸What they score: calibrated risk-taking and clear reasoning, not perfect arithmetic.

Quant research (systematic funds)

The emphasis is statistical rigour and honest research. Expect a take-home that is statistics plus a small data-analysis exercise on a provided dataset, then a discussion of your design choices. Probability, conditional expectation, regression and a clean backtest are the core.

• ▸Take-homes resembling those at firms such as Two Sigma and Citadel: build a signal, evaluate it honestly, defend it.
• ▸Whiteboard probability/statistics: conditional expectation, distributions, estimators.
• ▸What they score: no lookahead, sensible evaluation, and clear thinking about what could be wrong.

Quant dev

The emphasis is engineering and numerical care. Expect coding (often C++ or Python), a pricing or simulation problem (Black-Scholes, Monte Carlo, the Greeks), and questions on performance and correctness.

How to prepare for the loop you're in

• ▸Trading: drill mental maths and play market-making games out loud with a friend.
• ▸Research: do full take-homes end to end and practise defending your choices.
• ▸Dev: implement a pricer from scratch and benchmark it; know your complexity.

What the take-home round is really testing

For research seats the take-home is the interview — the code is a means to a conversation about judgement. They want to see that you can frame a question, evaluate your own work honestly, and say what could be wrong with it. A modest result you can defend beats an impressive one you can't. The follow-up discussion, where you justify your choices and concede the weaknesses, is usually what decides the offer.

Red flags that quietly sink strong candidates

• ▸Lookahead in a backtest, or a model fit on the whole sample then evaluated on it.
• ▸A Monte Carlo or backtest number with no error bar and no robustness check.
• ▸Overfitting defended as 'robustness' — many parameters, tuned on one period, presented as universal.
• ▸Can't state the assumptions behind a model, or can't answer 'what would make this fail?'.
• ▸Code that doesn't run as submitted, or results that don't reproduce.