Citadel Securities Quantitative Research Interview Guide

Citadel Securities quantitative research interview process

The campus research process centres on live problem solving followed by deeper technical conversations with researchers. Citadel Securities puts the overall journey at around four to five weeks, with timing shaped by interview scheduling and team review.

Citadel Securities is the market making firm, while Citadel is a separate investment management business. Campus recruiting offers opportunities to speak with both, but the interview details here apply specifically to Citadel Securities quantitative research.

1. Application and research fit

   Choose the role that matches your training: quantitative research analyst, quantitative researcher or a specific research internship. Show concrete evidence of statistical analysis, predictive modelling and programming rather than relying on a list of mathematical courses.

   For each project on your CV, prepare the research question, your contribution, the validation method and the result that mattered.

2. First round: 45 to 60 minutes, remote

   The video interview combines programming, algorithms, problem solving and behavioural discussion. Citadel Securities sends a CoderPad link in advance, and its principal research languages are Python and C++, although other languages are accepted.

   Practise narrating how you choose a data structure, check assumptions and revise a solution after a hint.

3. Second round: three to five onsite interviews

   The campus onsite normally comprises three to five interviews of about 60 minutes each. The conversations combine technical and behavioural questions, including discussion of research interests and previous work.

   Build the stamina to analyse several different problems, then explain a research decision in detail without notes.

4. Team review and next steps

   Hiring managers review candidates for research team fit after the onsite stage. If several teams show interest, your recruiting contact helps coordinate the match.

   Citadel Securities targets a response within two weeks of the second round, so share any competing decision deadline with your recruiter early.

The sequence above is the firm's campus quantitative research path. Applicants with substantial prior research experience should put more preparation time into their specialised work, model assumptions and commercial applications.

Explore the official campus research interview process.

First round: live CoderPad interview

Expect to explain the algorithm before writing code, then test what you have written. A strong answer moves from a correct simple approach to the right complexity and handles cases that break a superficial solution.

EverythingQuant CoderPad practice benchmark

Complete one unfamiliar algorithm problem in 50 minutes while speaking aloud. Spend five minutes defining the inputs, 25 minutes implementing, 10 minutes testing difficult cases and 10 minutes discussing complexity and alternatives.

For a realistic task, take the quote alignment exercise in the worked practice section and implement it from a blank editor. Explain why a data record that arrives after the prediction time must never affect the answer.

Onsite interviews: research, statistics and model judgement

The onsite tests whether you can solve technical questions and defend how you do research. Prepare for changes of direction between coding, probability, statistical inference and your own specialist projects.

Statistics and modelling questions

Work through ordinary least squares assumptions, confidence intervals, conditional expectation and hypothesis testing without relying on a formula sheet. Be able to distinguish predictive association from a usable trading signal.

When discussing regression, start with the data generating process: What is the target, what information exists at prediction time, and how are the residuals related across observations? Then describe a chronological validation design and how you would test stability under a market regime change.

Research design and market applications

Citadel Securities researchers build signals, pricing models and strategies from large market data sets. That makes leakage, repeated hypothesis testing, transaction costs, latency, capacity and shifting regimes especially valuable areas for practice.

A useful prompt is to design a model for the next ten seconds of midprice movement. Explain which quote and trade events are available when the forecast is made, how to avoid overlapping train and test information, and how to judge the forecast after trading costs.

Discussion with several teams

Prepare two distinct research narratives, one about a result that worked and one about an idea that failed. For each, explain what you controlled, what you measured, the strongest alternative explanation and the next experiment you would run.

Give clear answers to questions about collaboration and motivation as well. Interviewers want researchers who can defend a technical choice while responding constructively to criticism.

Choose preparation for your research background

The core skills overlap, but the strongest examples to bring depend on your training. Pick the panel closest to the role you applied for and use its exercises to prepare your own project explanations.

These are three research applicant backgrounds, not three different employers. Quantitative trading and quantitative development have separate responsibilities and deserve their own preparation rather than being folded into the research interview process.

Technical subjects worth prioritising

The most useful syllabus follows the firm's research workflow: infer something from noisy data, test it honestly and turn it into a decision that survives real trading conditions. Practise each topic by explaining a worked example, not by memorising definitions.

For probability fundamentals, use Probability for Quants. For interview implementation practice, use the programming question library.

Four original quantitative research practice questions

These EverythingQuant problems mirror the reasoning skills used in research and technical interviews; they are original exercises, not claims about proprietary Citadel Securities questions. Try each problem before opening the solution and explain your assumptions aloud.

def align_quotes(quotes, decisions, max_age=2):
    if max_age < 0:
        raise ValueError("max_age must be nonnegative")

    result = []
    quote_index = 0
    last_quote = None

    for decision_time in decisions:
        while (quote_index < len(quotes)
               and quotes[quote_index][0] <= decision_time):
            last_quote = quotes[quote_index]
            quote_index += 1

        if (last_quote is not None
                and decision_time - last_quote[0] <= max_age):
            result.append(last_quote[1])
        else:
            result.append(None)

    return result


quotes = [(1, 100.0), (3, 101.0), (3, 101.5), (7, 102.0)]
assert align_quotes(quotes, [0, 3, 5, 8, 11]) == [
    None, 101.5, 101.5, 102.0, None
]
assert align_quotes([], [1, 2]) == [None, None]
assert align_quotes(quotes, []) == []
assert align_quotes([(4, 9.0)], [3, 4, 5], 0) == [
    None, 9.0, None
]

Explain a research project like a quantitative researcher

Choose a project you can defend at the level of its data, assumptions, implementation and failure cases. The strongest interview story moves from the hypothesis to a result that withstands scrutiny.

1. Define the target. State exactly what is predicted or estimated and when the information becomes available.
2. Establish the baseline. Explain the simplest relevant comparison and why your method improves on it.
3. Describe validation. Walk through temporal splits, leakage checks, uncertainty and robustness tests.
4. Translate the result. Show how the finding affects decisions after real execution costs or operational constraints.
5. Own the limitations. Identify the most plausible reason the result could fail and the next test you would run.

A model answer you can adapt

My project predicted short horizon price changes from quote and trade data. I restricted features to information available before each forecast, validated on later trading sessions and compared against a simple baseline. The initial effect weakened in volatile periods, so I isolated that regime, removed an unstable feature and checked whether the improvement survived execution costs.

That example is an illustration of a good research explanation, not an experience to claim as your own. Replace every detail with a real decision and measurement from your work, and be able to explain the mathematics beneath it.

Questions to rehearse about your own work

• Which observation would most strongly falsify your hypothesis?
• What would change if the errors were correlated or the data were nonstationary?
• How much of the improvement remained after multiple comparisons and costs?
• Which component was hardest to implement and what did you trade away to make it reliable?
• How would you detect that a live model had stopped working?

A focused two week preparation plan

Use this EverythingQuant plan when your next interview is approaching. It gives priority to programming and research explanations because the first campus round tests both directly.

Score each mock on correct reasoning, implementation, edge case testing and clarity rather than speed alone. For additional practice, use EverythingQuant's probability interview questions.

Frequently asked questions

Official sources and next steps

Read the employer's research process alongside the specific posting you applied for. Use the role descriptions to decide which projects and mathematical topics deserve the most attention.

For adjacent technical preparation, explore Probability for Quants and programming interview practice. For a different Citadel software engineering career path, use the dedicated software engineering guide.