G-Research uses a quantitative quiz before research interviews, while engineering hiring begins with a technical interview or remote test. The research track, machine learning track and engineering track assess different skills, so use the process for your intended role.
Choose your role
G-Research Quantitative Researcher Interview Process
Quantitative researcher and quantitative analyst applicants complete a general online quant quiz, usually followed by four one hour technical interviews and leadership interviews. One technical interview focuses on mathematics in depth. Your goal is to derive results, explain assumptions and discuss research decisions under questioning.
Stage 1: Online Quant Quiz
Solve quantitative aptitude problems after the application review to demonstrate mathematical reasoning and problem solving.
Stage 2: Technical Interviews
Work through mathematical questions in four one hour interviews, including a deeper mathematics discussion.
Stage 3: Leadership Interviews
Explain your research choices, intellectual independence and approach to working with other specialists.
Stage 1: Online Quant Quiz
The general quantitative aptitude quiz is the first assessment stage after a successful application. Focus on conditional probability, expectation, random processes, covariance, regression, combinatorics and mathematical puzzles. The G-Research sample quant paper contains questions on reversing regression, probability puzzles, random sticks, domino tilings and stochastic processes.
Practise reaching a defensible answer before checking with code. The sample paper is representative preparation, not the exact questions in a current live quiz. Use the time limit, question navigation and permitted tools stated in your assessment invitation.
Train probability for the quant quizEverythingQuant probability course covering conditional probability, expectation, covariance and challenging practice problems. Some lessons require Premium.Stage 2: Technical Interviews
The main research process typically includes four one hour technical interviews, with one dedicated to deeper mathematics. Prepare to build a proof or derivation at the board, explain each assumption and adapt when an interviewer changes the question. Statistics, programming and research questions should connect to the skills in your application.
Mathematics
Derive conditional expectations, covariance identities, orthogonal projections, matrix rank and eigenvalue results. Check boundary and limiting cases.
Statistics
Write a likelihood, compare estimators, explain bias and variance, and test what happens when observations become dependent.
Programming and experiments
Implement a clear baseline, justify complexity and explain tests that separate a real effect from leakage or random noise.
When a question becomes difficult, name the random variables or constraints, solve a simple case, derive the general result, then test it numerically. Practise changing one assumption at a time in response to follow up questions.
Stage 3: Leadership Interviews
After technical assessment, research applicants meet G-Research leaders. Choose one research project you can explain without slides: hypothesis, baseline, method, evidence, failure cases and what you would test next. Be prepared to defend a choice and revise it when presented with better data.
How to Prepare for Quantitative Research Interviews
Work through unseen problems rather than repeatedly reviewing familiar solutions. An EverythingQuant practice benchmark is ten questions in 50 minutes: three probability, two statistics, two linear algebra or regression, two discrete mathematics and one algorithmic question. This is a study drill, not G-Research's quiz duration or question count.
Aim for eight fully justified solutions and record the first incorrect assumption in each missed problem.
Focused practice drills
- Solve six unfamiliar probability or expectation questions and explain one aloud in two minutes.
- Derive two covariance or regression results; verify each against a numerical example.
- Derive two likelihood estimators and identify a violated assumption for each.
- Hold two 30 minute mock technical discussions. For each wrong answer, log the first incorrect assumption and a shorter valid method.
Use a mock research project to propose a predictive feature, choose a temporal holdout and identify leakage. A strong answer makes the experimental decision precise rather than just describing a model.
G-Research Quant Research Interview Questions
These three original EverythingQuant practice problems develop the mathematical reasoning used in quant research. They are not claimed to be past G-Research interview questions.
1. Bayesian reasoning with a rare signal
A rare market event occurs on 10% of days. A detector raises an alert on 80% of event days and on 20% of other days. Given an alert, what is the probability that the event actually occurred?
Worked solution
The probability of an event and alert together is 0.10 × 0.80 = 0.08. The probability of an alert without the event is 0.90 × 0.20 = 0.18.
Bayes' rule gives 0.08 / (0.08 + 0.18) = 4/13 ≈ 30.77%. An 80% detection rate does not imply an 80% positive predictive value when the underlying event is rare.
Follow up: if the false alarm rate falls to 5%, the posterior becomes 0.08 / (0.08 + 0.045) = 64%.
2. Reversing a regression
Let X have mean zero and variance 4. Set Y = 3X + E, where E is independent of X, has mean zero and variance 9; find the population least squares slopes of Y on X and X on Y.
Worked solution
Independence gives Cov(X,Y) = 3 Var(X) = 12. Also, Var(Y) = 9 Var(X) + Var(E) = 45.
The slope predicting Y from X is 12/4 = 3. The reverse slope is 12/45 = 4/15, so reversing the variables does not invert the original slope.
Their product is 3 × 4/15 = 4/5, equal to the squared correlation. This identity is a useful check on your derivation.
3. Estimating a mean when observations have different noise
Three independent Gaussian measurements of a common unknown mean are 2, 3 and 5, with known variances 1, 4 and 1. Find the maximum likelihood estimate of the mean and its variance.
Worked solution
The Gaussian likelihood assigns each observation weight equal to the inverse of its variance. The weights are 1, 1/4 and 1, with total weight 9/4.
The weighted estimate is (2 + 3/4 + 5) / (9/4) = 31/9 ≈ 3.444. Its variance is 1 / (9/4) = 4/9, giving standard deviation 2/3.
Follow up: equal weighting would waste information because the middle observation has four times the noise variance of either other observation.
G-Research Machine Learning Researcher Interview Process
Machine learning researcher applicants follow a research hiring path that begins with a quantitative quiz selected for their background. The ML technical route includes two one hour interviews focused on machine learning, with mathematics, statistics and programming questions relevant to that work, followed by leadership interviews.
Stage 1: Online Quant Quiz
Complete the ML focused or general assessment assigned for your background, demonstrating quantitative reasoning.
Stage 2: ML Interviews
Defend models, methods and experiments during two one hour technical discussions with supporting maths and programming.
Stage 3: Leadership Interviews
Explain the contribution, reliability and practical limits of your strongest research work to leaders.
Stage 1: Online Quant Quiz
The research recruitment process has a general quantitative quiz and an ML specific version. The version depends on your background. For an ML route, revise estimation, optimisation, linear algebra, probability, model evaluation and the mathematical principles behind common learning algorithms. Train yourself to justify a loss function and explain how changing the data or metric changes the answer.
For practice, derive a gradient and a regularisation term by hand, then verify them with a minimal Python experiment. Treat those exercises as preparation topics rather than a confirmed list of exact quiz questions; follow the timing and question navigation rules in your invitation.
Revise probability before the ML quant quizEverythingQuant conditional probability, distributions and covariance exercises. Free introductory lessons and additional Premium content.Stage 2: ML Interviews
For the ML research route, the firm describes two one hour technical interviews. Explain the objective, optimisation algorithm, regularisation and evaluation method behind a model you understand deeply. Questions can extend into statistics, mathematics and programming, so rehearse a calculation or implementation rather than relying on terminology alone.
Discuss a model under changing conditions
Start with a baseline and a time based validation split. Explain how you detect overfitting, information leakage, dependence between observations and a shift in the prediction target. If interviewers challenge your result, propose an ablation or holdout experiment that could disprove your claim.
Stage 3: Leadership Interviews
Research leadership discussions follow the technical interviews. Describe an ML result you can reproduce and a result that failed; show the evidence that changed your modelling decision. Where a research paper is central to your application, explain exactly what you implemented and which claims you tested yourself.
How to Prepare for Machine Learning Interviews
Prepare a compact project defence: objective, baseline, split, feature construction, validation metric, leakage audit and experiment ablation. Pick one model whose gradient or update rule you can derive and whose failure cases you can reproduce in code. Practise answering the question, “What changes when the samples are dependent?”
As an EverythingQuant drill, implement a minimal algorithm, audit a mistaken train/test split and conduct a 30 minute mock research discussion in which the other person repeatedly alters the assumptions.
G-Research Machine Learning Interview Questions
This original EverythingQuant Python exercise trains leakage safe feature construction. It is an independent practice problem rather than a confirmed employer question.
4. Constructing a time series feature without leakage
At observation i, construct the average of the preceding k measurements using only observations before i. Return None when fewer than k previous values exist.
Worked Python solution
def previous_means(values, k):
if k < 1:
raise ValueError("k must be positive")
result = []
running = 0
for i, value in enumerate(values):
if i >= k:
result.append(running / k)
running -= values[i - k]
else:
result.append(None)
running += value
return result
assert previous_means([10, 12, 9, 15, 11], 3) == [
None, None, None, 31 / 3, 12.0
]
At index 3, the previous values are 10, 12, 9, giving 31/3. At index 4, they are 12, 9, 15, giving 12.
The current measurement enters the running total only after the feature is calculated. The algorithm uses O(n) total time and O(n) output memory, with constant extra working space.
G-Research Engineering Interview Process for Software Engineers and Quantitative Developers
Software engineers, research engineers, quantitative developers and machine learning engineers use the engineering recruitment route, not the two interview ML researcher route. The firm lists an initial technical interview or remote test, a behavioural interview, further technical interviews and management interviews. The relevant engineering speciality changes the technical questions.
Stage 1: Technical Screen
Show coding and engineering fundamentals in an interview with a team member or a remote technical test.
Stage 2: Behavioural Interview
Discuss personal ownership, collaboration and how you respond when projects or requirements change.
Stage 3: Technical Interviews
Go deeper into implementation, architecture, performance and specialist problems relevant to the team.
Stage 4: Management Interviews
Meet future teammates, your manager and functional leaders to discuss decisions and technical judgement.
Stage 1: Technical Screen
The engineering process starts with a technical interview with a team member or a remote test. Practise readable code with arrays, hash maps, queues, sliding windows and basic algorithms; test empty input, duplicate events, exact boundary times and malformed records. The precise programming language, platform, timer and ability to revisit questions depend on the supplied task and are not fixed by the engineering assessment process.
For a quantitative developer vacancy, G-Research identifies C# and Python as a core stack. Code to a correct input/output contract first; then discuss the runtime, memory and possible failure modes.
Practise programming interview questionsEverythingQuant coding exercises in streaming data, sliding windows, dictionaries and algorithms; some questions are Premium.Stage 2: Behavioural Interview
G-Research separates a behavioural conversation from its engineering technical rounds. Bring one specific example of a system you improved or a decision you changed after feedback. Explain the tradeoff, what you owned, who was affected and what you measured.
Stage 3: Technical Interviews
Later technical discussions go deeper into engineering competence. Quantitative developers build research tooling, simulation and production systems; research engineers work with large scale compute and modelling workflows. Be ready to explain data structures, profiling evidence, testing strategy, software reliability, concurrency or performance decisions relevant to your experience.
Quantitative development
Explain a simulation or backtest's state, failure recovery, reproducibility and the difference between faster code and a correct result.
Machine learning engineering
Describe training performance, data movement, efficient operations, memory constraints and the profiling evidence behind an optimisation.
Infrastructure and systems
Reason about a slow stream, a dropped message or a failed node. Explain monitoring, tests and recovery before discussing optimisations.
For coding questions, explain why a two pointer solution is linear rather than simply giving a complexity label. For a design question, identify the workload and failure assumptions before proposing components.
How to Prepare for Engineering Interviews
Solve two unfamiliar coding problems, add tests and defend complexity. For a streaming task, work through duplicates, timestamp ties, an empty sequence and the precise interval boundary. Then practise a project walkthrough that begins with what the software needed to accomplish, not the technologies you used.
The EverythingQuant Data Structures and Algorithms course covers queues, hash maps, sorting, graphs and algorithmic techniques. Its introductory content is free and additional lessons require Premium.
Stage 4: Management Interviews
The final engineering stage involves future teammates, your manager and functional leadership. Discuss one hard production decision with concrete evidence: a measured bottleneck, the implementation you chose, how you assessed reliability and the result for users of the system.
G-Research Software Engineering Interview Questions
This EverythingQuant problem tests a streaming window and the ability to verify endpoint conventions. It is a practice task, not a claimed G-Research interview question.
5. Finding the busiest event window
A system receives events with timestamps sorted in ascending order. Write an algorithm that finds the maximum number of events in any window of width 60 seconds, excluding the left endpoint and including the right endpoint.
Worked Python solution
def max_events_in_window(times, width):
if width <= 0:
raise ValueError("width must be positive")
left = 0
best = 0
for right, current in enumerate(times):
while times[left] <= current - width:
left += 1
best = max(best, right - left + 1)
return best
assert max_events_in_window([0, 25, 60, 61, 120], 60) == 3
assert max_events_in_window([], 60) == 0
Keep two pointers to the earliest and latest events within the current interval. Each pointer moves forward at most once per event, so the runtime is O(n) with O(1) extra memory.
The busiest window ends at second 61 and contains events at 25, 60 and 61. The window definition matters because an event exactly 60 seconds earlier is excluded.
G-Research Research and Engineering Internship Interviews
The two internship selection processes have different assessments, separate from the full time role stages above. G-Research lists a typical one to two week selection process for both.
Quantitative research and ML internships
Apply online, complete a quantitative aptitude quiz, take technical interviews and finish with a functional lead interview. For research internships the online stage may be the general quant quiz or ML version, followed by technical discussions with potential project mentors. Prepare probability, estimation, experimental design and a concise defence of an academic research project.
Software and engineering internships
Apply online, complete the engineering online skills test and attend an assessment centre. Practise correct code under time pressure, explain complexity, add edge case tests and describe one engineering project you completed yourself. Follow the assessment instructions in your invitation and practise accurate implementations under time pressure.
Spring into Quant Finance is an educational programme with a separate selection process, not an engineering or research internship interview.
Explain Your Research or Engineering Project
A project discussion is a chance to demonstrate reasoning you own. As an EverythingQuant practice exercise, prepare a ten minute walkthrough followed by at least twenty minutes of challenging questions about your assumptions, tests and results. Research candidates should emphasise validity and uncertainty; engineers should emphasise profiling, reliability, ownership and recovery.
| Interview prompt | What a good answer establishes |
|---|---|
| Why did your model improve? | A fair baseline, holdout design, metric, leakage checks, uncertainty and a useful ablation. |
| What if observations are dependent? | How dependence changes effective information, standard errors and validation. |
| How did you accelerate an experiment? | The original bottleneck, profiler evidence, change and reproducible before and after measurements. |
| Tell us about a result that failed. | The original hypothesis, diagnostic experiment, revised decision and lesson for the next test. |
Bring two concise examples of judgement and collaboration, including a time you changed your conclusion after criticism or new evidence. Specific results and decisions are more persuasive than a generic claim to work well in teams.
Start practising on EverythingQuantCreate a free account to explore interview questions, online assessments and preparation guides.G-Research Interview FAQs
Do I need a finance degree for quantitative research?
No. G-Research welcomes strong mathematical, physics, statistics, computer science and related technical backgrounds for research roles without requiring prior finance experience.
Do all applicants take the same online assessment?
No. Research applicants take a general or ML focused quant quiz, engineering applicants have a technical interview or remote test, and engineering interns take an online skills test.
How many technical interviews are in the research process?
The main quantitative research route typically uses four one hour technical interviews. The ML research route typically uses two one hour interviews that include ML with relevant mathematics, programming and statistics.
How do research and engineering internships differ?
Research internship selection includes a quant test, technical interviews and a functional lead interview. Engineering internship selection uses an online skills test and assessment centre.
Are these real G-Research interview questions?
The five worked examples are original EverythingQuant practice questions. The separate employer sample quant paper contains questions selected by G-Research for candidate preparation.
Official G-Research Interview Resources
These G-Research pages distinguish the research, engineering and internship routes, with a sample quant paper for independent practice.
- Quantitative research and ML team assessment process
- Quant interview questions and preparation guidance
- Internship assessment routes and student opportunities
- Research internship application and technical interviews
- Engineering assessment process
- Research engineer assessment and interview stages
- Quantitative developer role and technology
- G-Research sample quant paper