Derivatives Foundation interview preparation

Questions

100

Traced to a firm

29

Firms

19

Updated

September 2026

Asked at

All firmsMSMorgan Stanley4Nomura4Akuna Capital2Amundi2HSBC2PIMCO2Bank of America1Barclays1Citadel1DRW1Goldman Sachs1Jane Street1Millennium Management1Mizuho1Old Mission Capital1RCRBC Capital Markets1Scotiabank1UBS1Wells Fargo Securities1

Topic

All topicsForwards and futures10Options basics8Option pricing7The Greeks10Volatility7Option strategies9Swaps and rates7Credit derivatives4Market structure and clearing6Indian derivatives8Trading and markets9Brainteasers6Fit9

Level

AnyCoreIntermediateHard

Type

AnyTechnicalCaseMarket viewBrainteaserFit

1. 098When would you use C++ over Python on a trading desk?FitIntermediatetechnicalWells Fargo SecuritiesSales and Trading · Charlotte · 2026

   Say this

   C++ where latency and deterministic performance matter — the execution path, the pricing engine inside a market-making loop, anything in the hot path measured in microseconds. Python everywhere else: research, calibration, backtesting, reporting and glue. Most desks run both, with the Python layer calling into C++ libraries.

   Then walk it

   1. The real distinction is the hot path versus everything else. If code runs a million times a second between a market data tick and an order, you need C++ or something comparable — predictable memory layout, no garbage collector pauses, control over allocation and cache behaviour.
   2. Python's cost is not just raw speed, it is the interpreter's unpredictability. A garbage collection pause of a few milliseconds is irrelevant in research and fatal in a quoting engine.
   3. Where Python wins decisively: iteration speed in research, the numerical and data ecosystem, and readability for code that a team has to reason about. Pandas, NumPy and the scientific stack mean a volatility surface calibration takes an afternoon rather than a week.
   4. The standard architecture is both. Numerical kernels in C++ exposed through bindings, with Python orchestrating. The heavy libraries you use in Python are C or C++ underneath anyway, so a vectorised NumPy routine can be within a small factor of hand-written C++.
   5. So the practical answer: write it in Python first, measure, and only rewrite the part that is actually the bottleneck in C++. Premature optimisation in C++ costs weeks of development time for latency nobody was going to notice.
   6. And I would be honest about my own level. If I am stronger in Python, I say that and say what I have done in C++ — knowing enough to read a pricing library and understand why it is written the way it is, versus being able to write a low-latency engine, are different claims and I would not blur them.

   Where candidates lose it

   Answering 'C++ is faster' and stopping. The interviewer wants the hot-path distinction, the point that most production stacks are both, and the discipline of measuring before rewriting. And do not overclaim your C++ — a technical interviewer will test it in the next question.

   Expect next

   • What specifically would you put in C++ on a market-making system?
   • How fast can Python get if you vectorise properly?
   • How strong is your C++, honestly?

   Reported by candidates at Wells Fargo Securities (Sales and Trading, Charlotte, 2026). Source: Wall Street Oasis.

Firm tags come from public, anonymous candidate reports on Wall Street Oasis: strong signal, not sworn testimony. Firms are named as the places a question was reported, not as partners of Fin Maverick. Answers are written for this page to show how to think out loud; they are not scripts to recite.