The Best-Paying Quant Job Most UK Students Have Actually Heard Of
Every October, a G-Research stand appears at careers fairs in Cambridge, Oxford, Imperial and Warwick, usually running a maths puzzle competition with a cash prize, and every year a fresh cohort of mathematicians discovers that a London firm they cannot quite describe pays new PhD researchers a reported £150,000 to £300,000+ in year one. The firm behind the stand is one of the more unusual employers in UK finance: a quantitative research house, widely reported as part of the Trenchant group of entities, that develops machine-learning driven forecasting for financial markets and says very little else about itself.
G-Research occupies a specific niche in the UK quant landscape. It is London-first in a way that Citadel and Jane Street are not: the research organisation is centred in London, the graduate pipeline is built around UK universities, and the pay is calibrated to be exceptional in sterling terms rather than an FX translation of a New York package. A US engineering hub in Dallas has also been reported, but the centre of gravity is unambiguous.
This guide covers what G-Research actually does, the roles it hires for, the famous puzzle-heavy interview process with worked examples, reported compensation, and the confidentiality culture that anyone considering an offer should understand before signing.
G-Research at a Glance
- What it is: A London-based quantitative research firm, widely reported as linked to the Trenchant group
- Focus: Machine learning and statistical research for forecasting financial markets
- Offices: London (research and engineering), with a Dallas engineering hub reported
- Known for: Very high UK pay, the monthly puzzle, sponsorship of maths and CS events, and strict confidentiality
- Roles it hires: Quantitative Researcher, Machine Learning Researcher, Software Engineer, Data Engineer, Infrastructure Engineer
- Typical backgrounds: Maths, physics, CS and engineering from strong UK and European universities; many researchers hold PhDs
What G-Research Actually Does
G-Research describes itself as a research firm rather than a fund: teams of researchers and engineers build predictive models for financial markets using large datasets and machine learning, supported by a substantial internal computing platform. Reporting over the years has linked the firm's research to trading carried out within the wider group structure it is associated with.
For a candidate, the practical picture is this. Researchers work on forecasting problems: alpha signals, statistical models, ML architectures applied to noisy, non-stationary financial data. Engineers build the platform those researchers depend on, from data pipelines and GPU clusters to backtesting and experiment-tracking infrastructure. The firm has been public about investing heavily in large-scale compute, and its engineering job adverts read like a technology company's.
The work is closer to an industrial ML lab than to a trading floor. There are no traders shouting at screens; the output of a good week is a model that predicts slightly better than the previous one. People who want market adrenaline should look at prop firms; people who want research depth with financial-sector pay tend to thrive here.
Roles
Quantitative Researcher. The core role. Statistical modelling and signal research on financial datasets. PhD-heavy intake, though exceptional MSc and undergraduate candidates get through. Day-to-day skill mix: probability and statistics first, Python second, ML third.
Machine Learning Researcher. Deeper on modern ML: architectures, large-scale training, optimisation. G-Research sponsors NeurIPS, ICML and similar conferences and recruits from that community.
Software Engineer. Platform, data and infrastructure work at serious scale. C#, Python and C++ have all featured in the firm's public adverts over the years, alongside substantial Kubernetes and GPU-cluster work.
For how the researcher role compares across firms, see our quant researcher salary guide.
The Interview Process
Candidates typically report three to four stages over three to six weeks. The exact loop varies by role; this is the researcher-track shape.
Stage 1: Online Assessment
A timed test of mathematics, statistics and logic. The style is distinctive and worth practising for: probability brainteasers, combinatorics, and quick quantitative reasoning, pitched harder than most banks' tests and comparable to top prop firms'. Engineers get an algorithmic coding assessment instead.
Stage 2: Technical Phone Screen
A 45 to 60 minute call with a researcher or engineer. Researchers face probability and statistics questions solved live, often with follow-ups that generalise the problem. Expect classic material done properly: conditional probability, expectation, distributions, estimators and their properties.
Stage 3: Onsite
A half or full day in London. Several sessions covering harder probability and statistics, ML fundamentals for the research tracks (bias-variance, regularisation, why deep models overfit financial data so readily), a discussion of your own research, and coding. Puzzle-style questions in the spirit of the firm's public monthly puzzle appear throughout the process.
Stage 4: Offer
Usually within a week or two. G-Research moves decisively for candidates it wants, and packages for strong PhD candidates are among the best first offers in the UK in any industry.
Real Question Types
Probability
Question 1: The last passenger. 100 passengers board a plane with assigned seats. The first sits in a random seat; each subsequent passenger takes their own seat if free, otherwise a random free seat. What is the probability the last passenger gets their own seat?
Approach: 1/2. At every point where a displaced passenger chooses randomly, the choice that matters is between seat 1 and seat 100, and by symmetry each is equally likely to be taken first. Interviewers want the symmetry argument, not a simulation. This and its relatives are covered in our probability interview questions.
Question 2: Biased coin estimation. You flip a coin of unknown bias 20 times and see 14 heads. Give an estimate of the bias, a measure of uncertainty, and describe how you would test fairness.
Approach: MLE is 0.7; the standard error is sqrt(p(1-p)/n) ≈ 0.10, so a rough 95% interval is 0.5 to 0.9. A two-sided binomial test of p = 0.5 gives a p-value around 0.115, so you cannot reject fairness at 5%. The question is a statistics literacy check: estimator, uncertainty, test, in that order.
Statistics and ML
Question 3: Overfitting in finance. Why do ML models that work in vision and language often fail on financial prediction?
Approach: Signal-to-noise is orders of magnitude lower, the data-generating process is non-stationary (the market adapts to exploited patterns), effective sample sizes are small once you account for cross-sectional and serial correlation, and the researcher's own iteration introduces selection bias. Strong answers mention multiple-testing corrections and honest out-of-sample discipline.
Puzzles
Question 4: The 100 lockers. 100 lockers, all closed. Person k toggles every kth locker, for k from 1 to 100. Which lockers end open?
Approach: A locker is toggled once per divisor of its number, so it ends open when it has an odd number of divisors, which happens exactly for perfect squares. Lockers 1, 4, 9, ..., 100. Fast pattern recognition on divisor-counting arguments is very much in the G-Research house style; the Green Book drills dozens of problems in this family.
Coding
Question 5: Streaming median. Design a structure that supports inserting a number and querying the median, both efficiently.
Approach: Two heaps: a max-heap for the lower half and a min-heap for the upper half, rebalanced so their sizes differ by at most one. Insert is O(log n), median is O(1). Standard, but expected to be produced cleanly and quickly; our coding interview questions cover the wider repertoire.
Compensation: Among the Best in the UK
G-Research publishes no pay data, but reported offers have been consistent for several years across levels.fyi, Glassdoor and candidate accounts.
| Role | Level | Reported total package (London) |
|---|---|---|
| Quantitative Researcher | New PhD / graduate | £150,000 - £300,000+ year 1 |
| Quantitative Researcher | Experienced | £300,000 - £700,000+, performance-dependent |
| Machine Learning Researcher | New PhD | £150,000 - £300,000+ year 1 |
| Software Engineer | Graduate | £70,000 - £120,000 year 1 |
| Software Engineer | Senior | £120,000 - £250,000+ |
Packages are typically structured as base plus a generous discretionary bonus, with sign-on payments reported for strong candidates. The striking fact is the sterling-denominated level: for a new PhD staying in the UK, reported G-Research offers have often beaten everything except the US trading firms' London outposts. For context across the market, see our UK quant salary guide.
Culture and Confidentiality
Two things define the G-Research employment experience by reputation, and candidates should weigh both.
Research culture. Reasonable hours, serious compute, strong colleagues, and a genuinely academic flavour: the firm sponsors conferences, funds PhD programmes and runs public challenges. For a certain kind of mathematician it is close to an ideal industrial home.
Confidentiality. G-Research is strict about protecting its intellectual property, with tight information controls internally and restrictive covenants in contracts. The firm has been involved in widely reported legal disputes with former employees over confidential information. None of this is unusual in kind for the quant industry, but the intensity is at the high end, and anyone signing should read the non-compete and garden-leave terms carefully and price them into the decision. A large package looks different when leaving involves a long paid pause before you can join a competitor.
Where This Guide Is on Thin Ice
Most of what is publicly known about G-Research comes from its own recruiting materials, press reporting and candidate accounts, and the firm confirms little. The corporate structure behind the research house is reported rather than disclosed in any detail, compensation figures are estimates with wide error bars, and interview loops change year to year. The confidentiality picture above is based on reporting of past disputes, not on the firm's current contracts, which we have not seen. Verify terms directly before making decisions on them.
Compensation & recruiting notes
Pay figures in this guide are illustrative estimates from publicly discussed bands and anecdotal reports - not official figures from the employer. Packages vary widely by role, performance and year. Hiring processes change; nothing here guarantees an interview, assessment format or offer.
Frequently Asked Questions
What does G-Research do?
G-Research is a London-based quantitative research firm that builds machine-learning and statistical models for forecasting financial markets. It is widely reported as part of the Trenchant group of entities, and its research supports trading carried out within that wider structure.
How much does G-Research pay?
Reported packages for new PhD quantitative researchers are roughly £150,000 to £300,000+ in year 1, typically as base plus a generous discretionary bonus, with experienced researchers reportedly earning substantially more. These are estimates from public and anecdotal sources rather than employer figures.
How hard is the G-Research interview?
Hard, in a specific way: heavy on probability, statistics and puzzle-style mathematics, with ML depth for research roles. The online assessment filters aggressively, and the onsite includes live problem-solving with layered follow-ups. Olympiad-style comfort with brainteasers helps more here than at most hedge funds.
Do I need a PhD to join G-Research?
For the researcher tracks a PhD is the norm but not an absolute rule; exceptionally strong MSc and undergraduate candidates are hired. Engineering roles do not require one.
What is the G-Research non-compete situation?
The firm is reportedly strict on confidentiality and restrictive covenants, and has been involved in publicised disputes with former employees. Specific terms vary by contract and change over time, so read your own agreement carefully and take advice before signing rather than relying on reputation in either direction.
How does G-Research compare with Jane Street or Citadel in London?
Pay at the top of G-Research's reported researcher bands is competitive with the US trading firms' London offices, and the work is more research-lab than trading-floor. The main differences are asset-class breadth, culture and the confidentiality regime. Candidates who want market-facing roles usually prefer the trading firms; candidates who want pure modelling depth often prefer G-Research.
Practise the questions G-Research: Careers, Salary & Interview Guide 2026 actually asks
Reading about the interview is one thing - sitting one is another. Open your free Quantt prep workspace for a real course lesson plus interview-style coding tests modelled on firms like Jane Street, Citadel, Hudson River and Optiver.
Free lesson + interview practice · No credit card required