XTX Markets has just over 300 people, trades about $250 billion a day, and has no open graduate programme on its job board. The thing most applicants get wrong is waiting for a campus deadline that never comes. XTX hires a small number of people into specific research and engineering roles all year, and its engineering team openly says most of its staff are self-taught. What gets you in is deep, provable skill in one area, not a target-school CV. This guide covers who they hire, how the process runs, and what gets people cut.
Who they actually hire
XTX was spun out of GSA Capital in January 2015 by Alex Gerko (Wikipedia, retrieved September 2026). Its job postings describe price forecasts for over 53,000 instruments, over $250 billion traded a day across 35 countries, and over 300 employees in London, Singapore, New York, Paris, Bristol, Mumbai, Yerevan and Kajaani (XTX Software Developer posting, updated 17 September 2026). Young and Calculated's August 2026 profile puts 2025 net revenue at £3.93 billion, up 44%, and revenue per employee at about £13.1 million.
The firm is expanding on compute, not headcount. The same postings describe a research cluster of well over 25,000 GPUs with more than an exabyte of storage, and a data-centre build in Finland. Young and Calculated reports an investment of over €1 billion in Kajaani, with the first facility at 22.5MW. The core UK entity grew from 113 to 127 employees in 2025 (eFinancialCareers, April 2026). Fourteen net hires in a year at the main company tells you how few seats there are. Its outside spending points the same way: on 16 September 2026 XTX closed its AIMO maths prize, saying open-weight models now get perfect scores on olympiad problem sets, and moved $10 million each to Lean FRO and the Mathlib Initiative (XTX news). The firm spends on AI and mathematics, not on a large graduate class.
Estimates of the annual intake agree that it is small. Quantt's May 2026 guide says roughly 30 to 50 researchers and engineers a year worldwide. Quant Blueprint's 2026 guide says only a handful. What each track asks for is spelled out in XTX's own postings:
| Track | Who XTX asks for | Source |
|---|---|---|
| Quant research | Mostly PhDs in ML, statistics, physics; strong master's candidates with industry experience considered | Quant Blueprint, 2026 |
| Exchange trading development (C++) | Modern C++, CPU internals, algorithms; "most of our existing team are self-taught", CS degree not required | XTX C++ Software Engineer posting, Singapore, Aug 2026 |
| Platform / shared engineering | Go, build systems, CI/CD, developer tooling, including AI coding-assistant tooling | XTX Software Developer posting, London, Sep 2026 |
| ML performance | C/C++/Rust, GPUs/TPUs, AI compilers, kernel work | XTX ML Performance Engineer posting, London, 2026 |
| XTY AI Research Internship | PhD in progress with 1+ year left, ML publications, transformers; $35,000/month base | XTX Greenhouse, retrieved 27 Sep 2026 |
| Junior quant researcher, Yerevan | High-GPA maths/physics/CS graduates; olympiad medals desirable; English and Russian | XTX posting via Built In, removed Mar 2026 |
The Yerevan role is the closest thing XTX has had to a graduate scheme recently: an online test and then four or five interviews covering statistics, probability, general maths and algorithms. XTX said it wanted strong performance in all four areas and exceptional strength in at least one. That standard applies across the firm.
The process, step by step
Tradermath's July 2026 guide and Quantt's May 2026 guide describe the same shape:
| Stage | Format | What is graded | Timeline |
|---|---|---|---|
| Online assessment | 90-120 min. Engineers: 2-3 hard algorithm problems. Researchers: ML and probability paper with a data task | Correctness under time pressure; ~5-10% pass | Week 1-2 |
| Phone screen | ~60 min with a working researcher or engineer | Live coding or ML discussion; technical from the first minute | Week 2-3 |
| Super day | 5-6 interviews of 45-60 min, London | See split below | Week 3-5 |
| Decision | Written offer or rejection | Reapply after 12 months if rejected (Tradermath) | 1-2 weeks after |
Tradermath gives the super-day split. For researchers it is about 40% ML theory and practice, 25% probability and statistics, 25% Python and data coding, and 10% behavioural. For engineers it is about 50% coding, 30% systems design, 10% C++ specifics and 10% behavioural. It also reports a quick-fire C++ screen for some engineering teams: around twenty short questions in about thirty minutes on RAII, alignment, vtables, variadic templates, cache lines and floating point.
The XTY internship adds a research deep dive. Its posting lists what the interviews look for: your research experience, modern deep learning, classical ML, general maths (linear algebra, analysis, probability) and hands-on programming and data engineering.
What is not on the list matters too. Tradermath says XTX is not an options market maker like IMC, Optiver or SIG: there are no make-me-a-market dice games, very few classic brainteasers, and mental arithmetic speed is not what they are testing.
What gets people rejected
1. Knowing ML only at the surface. Quantt lists this as a main reason for rejection: candidates who can name a method but cannot derive it or explain when it fails. Its sample questions include pitfalls in time-series cross-validation and why walk-forward validation fixes them, and how to choose regularisation for a model that overfits.
2. Getting defensive when pushed. Quantt also names this. Interviewers keep adding constraints to see how you change your answer.
3. Preparing for the wrong firm. Candidates who drilled Optiver-style mental maths and market-making games arrive with nothing for a 40% ML super day (Tradermath).
4. Underestimating the engineering onsite. In an October 2025 Blind thread, a Bloomberg engineer who received an offer said the onsite had several LeetCode-hard problems plus system design and was a little harder than most FAANG and HFT loops. An Atlassian engineer who interviewed about ten years earlier as a new grad called it tougher than tech interviews and did not get an offer.
5. Knowing C++ without knowing the machine. XTX's C++ posting asks for caching, paging, system calls, out-of-order execution and how an indirect branch target is predicted. Knowing the syntax is not enough.
6. Waiting for a deadline. The board had nine open roles on 27 September 2026 and one student role. If you only check in September, you can miss the role that fits you.
A four-week plan
Week 1: pick one lane and make it provable. Researchers: put one reproducible project at the top of your CV, such as a time-series forecasting study with honest walk-forward evaluation. Engineers: one shipped system in modern C++ or Rust with measured latency numbers. Run the file through the free ATS résumé checker so the parser actually reads your skills. Set a weekly reminder to check XTX's Greenhouse board.
Week 2: build for the OA. Engineers: timed sets of two or three hard problems in 90 minutes, then review the constant factors. Researchers: conditional expectation, bias-variance, regularisation and a pandas data task against the clock. Tradermath's example: if X and Y are independent standard normals, E[X | X + Y = z] = z/2 by symmetry. Practise explaining why, not just giving the answer.
Week 3: go deep, out loud. Researchers: derive the methods on your CV, including the failure cases. Engineers: work through the C++ quick-fire list, then design a streaming statistics service and a lock-free queue and explain the memory ordering (both are on Quantt's sample list).
Week 4: mock super day. Five 45-minute sessions back to back, with someone pushing back on every answer. Also track the other London and New York firms with open 2027 roles on the New Grad 2027 tracker, because 30-50 seats a year is not a plan on its own.
Practice
XTX has no public question bank. The nearest published loops are its peers': Applr's Jane Street quantitative researcher mock asks follow-up questions on your probability reasoning, as XTX's phone screen does, and the DeepMind research engineer mock covers the ML depth that makes up 40% of XTX's research super day. For how XTX compares with the larger London employer, see XTX Markets vs Jane Street.
Visa notes
London roles use the Skilled Worker visa. GOV.UK sets the floor at £41,700 or the going rate, whichever is higher. New entrants can be paid 70% of the going rate with a minimum of £33,400 if they are under 26 or switching from a Student or Graduate visa, capped at four years in total. Every public estimate of XTX graduate pay is far above that floor. In New York, h1bdata.info shows XTX Markets LLC filing H-1B applications, for example two Trading Analyst filings at $110,000 in January 2022, so the lottery applies there after OPT.
Sources
- XTX Markets, Greenhouse job board, retrieved 27 September 2026
- XTX Markets, AI Research Internship – XTY Labs posting, retrieved 27 September 2026
- XTX Markets, C++ Software Engineer posting (Singapore), updated August 2026
- XTX Markets, Software Developer – Shared Engineering posting (London), updated 17 September 2026
- XTX Markets, Machine Learning Performance Engineer posting (London), 2026
- XTX Markets, "Update on XTX Markets' AI for Maths Philanthropy", 16 September 2026
- XTX Markets, Careers, retrieved September 2026
- Built In, XTX Markets Junior Quantitative Researcher (Armenia), removed March 2026
- Tradermath, "XTX Markets Interview Guide", 31 July 2026
- Quantt, "XTX Markets Interview", 9 May 2026
- Quant Blueprint, "How to Get a Job at XTX Markets in 2026"
- techinterview, "XTX Markets Interview Guide 2026", updated 21 July 2026
- Young and Calculated, "XTX Markets: The 300-Person Firm Trading 40x the London Stock Exchange", 2 August 2026
- eFinancialCareers, "Pay at XTX Markets rose modestly", 7 April 2026
- Blind, "XTX Markets Interview Experience - Software Engineering", October 2025
- h1bdata.info, XTX Markets LLC Trading Analyst filings, retrieved September 2026
- GOV.UK, Skilled Worker visa: when you can be paid less, retrieved September 2026
- Wikipedia, "XTX Markets", retrieved September 2026
Frequently asked questions
- How hard is it to get into XTX Markets?
- Very hard, mainly because there are so few seats. Quantt's May 2026 guide estimates XTX hires roughly 30 to 50 researchers and engineers a year worldwide and reports an online-assessment pass rate of around 5 to 10%. Tradermath's July 2026 guide gives the same pass-rate range and calls the assessment one of the hardest in the industry.
- Does XTX Markets have a graduate programme?
- Not an open one right now. On 27 September 2026 XTX's Greenhouse job board listed nine roles, and the only student role was the AI Research Internship at XTY Labs in New York, which is for PhD students available from January 2027. Graduates apply to specific engineering or research roles instead, so watching the board matters more than waiting for a campus deadline.
- Do you need a PhD to work at XTX Markets?
- For research, it helps a lot: Quant Blueprint's 2026 guide says the vast majority of XTX researchers hold PhDs, and the XTY internship asks for a PhD in progress with a publication record. For engineering, no. XTX's own C++ Software Engineer posting says most of the existing team is self-taught and a computer science degree is not necessarily a prerequisite.
- What does XTX Markets pay graduates?
- XTX does not publish graduate salaries, and the estimates disagree. Quantt gives £200,000 to £350,000 as an illustrative London graduate researcher package, while techinterview's July 2026 guide gives £100,000 to £200,000 first-year total pay for new-grad researchers and engineers. The only firm-published number is the XTY AI Research Internship base of $35,000 a month. The core UK entity paid an average of £457,000 per employee in 2025 (eFinancialCareers, April 2026), but that includes senior staff.
- Does XTX Markets sponsor visas?
- London roles use the UK Skilled Worker route, where GOV.UK sets a general floor of £41,700 or the going rate, and £33,400 for new entrants under 26 or switching from a Student or Graduate visa. In the US, h1bdata.info shows H-1B filings by XTX Markets LLC, for example two Trading Analyst filings in New York at $110,000 in 2022. Confirm sponsorship for your specific role with the recruiter.
- What is on the XTX Markets online assessment?
- Tradermath (July 2026) and Quantt (May 2026) both describe a 90 to 120 minute test. Engineers get two or three hard algorithmic problems; researchers get a combined machine learning and probability paper, often with a small data-analysis task. There are no market-making games, and Tradermath says rapid mental arithmetic is not the point.