Jump opened its 2027 campus roles region by region this autumn, starting with the US, and its own hiring page asks you to hold off on a US application if you intend to apply somewhere else. The mistake most applicants make is treating Jump as one application. It is eight separate campus tracks with different loops, different evidence and, in one case, a $300,000 annualised intern base — and choosing the wrong one produces a rejection nobody will explain to you. The second mistake is assuming Jump only wants C++ competitive programmers; its deep-learning research track asks for NeurIPS and ICML papers, not Codeforces rating.
Who they actually hire
Jump was founded in 1999 by Paul Gurinas and Bill DiSomma, two former Chicago Mercantile Exchange pit traders, and is still headquartered in Chicago with offices in New York, Austin, London, Amsterdam, Paris, Bristol, Singapore, Shanghai, Hong Kong, Mumbai, GIFT City and Sydney (Wikipedia, retrieved September 2026).
Size depends on who is counting. Revelio Labs' tracker puts Jump at roughly 2,357 people in early 2026, up about 25.5% from 2023, while Wikipedia lists 2,000+ (both retrieved 23 September 2026). Model-based trackers disagree by a couple of hundred, so read it as "about two thousand people and growing."
The more useful number from the same Revelio tracker is the direction of travel in postings: active Jump requisitions fell 6.8% in 2026 to 71, and monthly posting velocity dropped from 38 in 2023 to 17 in 2026. Headcount up, postings down and posted later means your timing matters more than your application volume. There is no benefit to spraying; there is a real cost to missing a window.
On selectivity, nothing official exists. Quantt estimates tens of thousands of applications against roughly 100 graduate hires globally a year. Quant Blueprint estimates acceptance "well below 5%" and describes the profiles Jump draws from: competitive programmers (ICPC, Codeforces, TopCoder), people with systems projects or open-source contributions, quant researchers with statistics and optimisation depth, and academic researchers. For the PhD research track, Jump's own campus material asks for a PhD in progress in statistics, mathematics, computer science, physics or any highly quantitative field.
The eight 2027 tracks, and picking the right one
Extern's guide, published 15 August 2026 and updated in September, lists eight campus tracks open for Summer 2027: Quantitative Trader, Quantitative Researcher (separate undergraduate/master's and PhD tracks), AI Research Engineer in three variants, Software Engineer, UI Software Engineer, Data Engineer, Systems Engineer and FPGA Engineer.
| Track | Aimed at | The evidence it wants | What to know |
|---|---|---|---|
| Quantitative Trader | Undergrad/master's, fast under pressure | Probability, expected value, market reasoning | 10-week programme; Tradermath describes a dedicated round on futures microstructure, options theory and HFT strategy |
| Quantitative Researcher (UG/MS) | Strong stats plus programming | Statistics, data mining, mathematics, machine learning | Phone screens go into probability, statistics, OLS and linear algebra (Tradermath) |
| Quantitative Researcher (PhD) | PhD in a quantitative field | Your own research, defended in depth | 10-week, in person, Chicago and New York, Summer 2027 |
| AI Research Engineer – Deep Learning | ML researchers | Publications at ICML, ICLR, NeurIPS, AAAI, UAI, KDD or equivalent OSS; transformers or SSMs; PyTorch, JAX or TensorFlow | $300,000 annualised base on the posting; bonus points for HPC, distributed training, CUDA or ROCm |
| AI Research Engineer (standard, Research Automation) | ML engineers and tooling people | Python and/or C++, end-to-end model development | Same family of postings, different emphasis |
| Software Engineer / UI / Data / Systems | CS and engineering | Algorithms plus systems depth | Quantt describes the online assessment as 90-120 minutes of two or three hard problems |
| FPGA Engineer | Hardware | Digital design, low-latency hardware | Smallest track; rarely covered in interview guides |
Two things fall out of that table. First, the AI research tracks are not a rebrand of software engineering — the deep-learning posting asks for a publication record and names the conferences, and pays a $300,000 annualised base for an intern (Jump posting, retrieved September 2026). If you have papers, that is the track where your evidence is worth the most. Second, if you do not have papers, applying there wastes the window you should have spent on quant research or engineering.
The AI emphasis is not cosmetic. Jump took an additional 11,000 square feet at One International Finance Centre in Hong Kong and is recruiting for quantitative trading and artificial intelligence roles there (Hedgeweek, 7 September 2026, following Bloomberg). Separately, its prediction markets team roughly doubled during 2026 to about 20 people, trading on Kalshi and Polymarket, with hires coming from sports betting and accounting rather than the usual quant pipeline (Bloomberg, 16 July 2026). Jump is hiring for shapes it did not hire for three years ago.
The process, step by step
| Stage | Format | What is graded | Timing |
|---|---|---|---|
| Online assessment | 90-120 min, two to three hard problems for engineering; a probability-and-statistics test for trader and research tracks (Quantt) | Correctness under a clock | Quantt estimates a 10-15% pass rate, labelled an estimate, not a disclosure |
| Phone screens | Two or three calls of 45-60 minutes (Quantt) | Traders: EV and brainteasers with market reasoning. Quants: probability, statistics, OLS, linear algebra. Developers: data structures and algorithms in C++ (Tradermath) | Roughly a week apart |
| Super Day | Four interviews (Tradermath) to five to seven (Quantt), 45-60 minutes each, usually Chicago | Two maths/quant rounds, one coding, one fit, with escalating difficulty | Offers typically within one to two weeks (Quantt) |
| Team match | Conversations with desks after the offer | Fit, not gating | Adds days, not weeks |
End to end: three to four weeks once underway (Tradermath), or four to five rounds across three to six weeks (Extern). If a headhunter sourced you instead of a campus application, Tradermath notes it can stretch into months. Quantt says you may reapply after 12 months.
The question set is consistent across guides. Tradermath names a card-stopping game — when do you stop calling the next card — as a recurring Jump staple, alongside coding prompts including a maximum-likelihood estimate from arrival times, ordinary least squares written from scratch, a work-stealing deque, and arbitrating an A and B market-data feed. Quantt adds order-book operations, cache-line awareness, lock-free data structures, adverse selection, and how you adjust a quote when the market moves against you.
If you want the firm-by-firm framing rather than the Jump-only view, the comparisons with DRW and Hudson River Trading cover pay and loop differences in detail.
What gets people rejected
Applying to the wrong region first. Jump's students and new grads page asks applicants to hold off on US positions if they are planning to apply in another region, because international roles are rolling out region by region (retrieved 23 September 2026). This is a published rule that costs you silently if you ignore it.
Solving correctly without saying anything. A Blind poster who interviewed for a core development role in July 2025 went through the online assessment, a phone screen and a three-hour onsite of three coding interviews, said the questions were not especially difficult, and did not advance — they attributed it to not explaining their thought process clearly enough. At a firm where the Super Day is mostly live problem solving with traders and engineers, silence reads as absence.
Preparing for the wrong assessment. A poster on Nowcoder who had just finished a Jump internship online assessment reported three questions and said they were not algorithm questions. Guides describe different OA shapes for different tracks — hard algorithmic problems for engineering, probability and statistics for trading and research — so confirm which one you are sitting before you spend three weeks on the wrong preparation.
Stopping at the first working solution. Quant Blueprint's failure list is specific: brute-force answers, poor or outdated C++, inability to optimise further when asked, and missing systems-level intuition about performance. The follow-up question "can you make it faster" is the interview, not an afterthought.
No firm-specific reasoning. Jump's roots are Chicago futures, and rounds go there. Being able to talk about how a futures contract behaves differently from an equivalent ETF, or why a quote should be pulled when flow turns informed, separates you from candidates who prepared generically for "quant".
What international students should know
Jump's Campus AI Research Engineer – Deep Learning posting says it plainly: international students are encouraged to apply, Jump accepts students eligible for CPT/OPT, and it sponsors work visas for full-time positions (retrieved September 2026). Most quant firms leave that unstated and make you ask; terms still vary by office and requisition, so confirm for the specific role, but the starting posture is favourable.
The region rule matters more for you than for a domestic applicant. If your plan is London, Amsterdam, Hong Kong, Singapore or Sydney — often the better plan if your work authorisation is easier there than in the US — then applying to a US requisition first is the wrong opening move, because Jump explicitly asks you not to. Hong Kong is the most active of those right now given the September office expansion, and the roles named there are quantitative trading and AI.
If you are running a US new-grad search alongside applications at home, the New Grad 2027 tracker lists what is open and when it closed, which is the part that catches people whose term dates differ from the US academic calendar.
A four-week plan
Week 1 — decide the track and the region, then apply. Pick one track based on the evidence you actually have: papers at named conferences for deep learning research, statistics and programming depth for quantitative research, speed and market reasoning for trading, systems depth for engineering. Decide your region before you submit anything. Read the posting's bonus qualifications and get them onto page one of your CV. Run your CV through a screen — Applr's free ATS checker shows what a parser extracts from your file, which is the version a recruiter sees first.
Week 2 — build for the OA you will actually sit. Engineering track: timed sets of two to three hard problems in 90-120 minutes, written in C++ where the role calls for it, with a second pass for constant factors every time. Trading and research tracks: conditional probability, expected value and Bayesian updating under a clock, plus OLS derived by hand rather than recalled.
Week 3 — Super Day rehearsal, out loud. Work the named prompts: the card-stopping game, maximum likelihood from arrival times, OLS from scratch, a work-stealing deque, arbitrating two feeds. Do them while narrating, because the July 2025 Blind account is the common failure, not the maths. Time-box each to 45 minutes.
Week 4 — firm-specific and fit. Learn Jump's actual business shape: Chicago futures roots, low-latency engineering, a growing AI research organisation, a prediction markets desk that doubled in 2026. Prepare two specific reasons this firm rather than the four other logos on your list, and a short, honest account of a decision you got wrong under pressure.
Practice
Jump does not have a public question bank, and the closest published loops are its peers'. Applr's HRT quantitative researcher mock runs the same probability-plus-research shape as Jump's research phone screens, and the Citadel quantitative researcher mock pushes on defending your own projects, which is where the PhD track lives or dies.
Sources
- Jump Trading, "Students & New Grads", retrieved 23 September 2026
- Jump Trading, Campus AI Research Engineer – Deep Learning (Intern) posting, retrieved September 2026
- Hedgeweek, "Jump Trading expands Hong Kong footprint", 7 September 2026
- Bloomberg via report, "Jump Trading expands prediction markets team as volumes surge", 16-17 July 2026
- Quantt, "Jump Trading Interview: Process", retrieved September 2026
- Tradermath, "Jump Trading Interview Guide", retrieved September 2026
- Quant Blueprint, "How to Get a Job at Jump Trading in 2026", retrieved September 2026
- Extern, "Jump Trading Internship 2027–2028: Deadlines & How to Apply", 15 August 2026, updated September 2026
- Revelio Labs, Jump Trading employee count and postings, retrieved 23 September 2026
- levels.fyi, Jump Trading Quantitative Researcher compensation, retrieved 23 September 2026
- Blind, "jump trading interview", July 2025
- Nowcoder, Jump Trading internship OA thread, retrieved September 2026
- Wikipedia, "Jump Trading", retrieved September 2026
Frequently asked questions
- How hard is it to get into Jump Trading?
- There is no official acceptance rate. Quantt's guide estimates Jump receives tens of thousands of applications and hires roughly 100 graduates globally each year, and Quant Blueprint estimates acceptance is well below 5%. Both are third-party estimates built from candidate reports, not Jump disclosures. What is verifiable is that Revelio Labs' tracker shows active Jump postings down 6.8% in 2026 to 71, with monthly posting velocity falling from 38 in 2023 to 17 in 2026 (retrieved 23 September 2026).
- Should I apply to Jump Trading's US roles or wait for my region?
- Jump's own students and new grads page says it is opening 2027 roles by region, starting with the US, and explicitly asks applicants to hold off on applying to US positions if they plan to apply to a role in another region (retrieved 23 September 2026). If you want London, Amsterdam, Hong Kong, Singapore, Sydney, Paris, Bristol, Mumbai, GIFT City or Shanghai, wait for that region to open rather than firing off a US application first.
- What does Jump Trading pay its interns?
- It depends entirely on the track. Jump's Campus AI Research Engineer – Deep Learning intern posting lists a $300,000 annualised base salary (Chicago or New York, retrieved September 2026). Extern reports software engineering interns at up to $120 an hour in Chicago plus a $20K sign-on and corporate housing (15 August 2026). For full-time, levels.fyi's Jump quantitative researcher median is $280,000 — $175K base plus $105K bonus, no equity, from 12 US submissions (retrieved 23 September 2026).
- Do I need to be a competitive programmer to get into Jump?
- For the engineering tracks it helps — Quant Blueprint lists ICPC, Codeforces and TopCoder backgrounds among the profiles Jump draws from. But the deep-learning research track asks for something different: a publication record at ICML, ICLR, NeurIPS, AAAI, UAI or KDD or equivalent open-source work, familiarity with transformers or state-space models, and PyTorch, JAX or TensorFlow (Jump posting, retrieved September 2026). Pick the track that matches your evidence.
- Does Jump Trading sponsor visas for international students?
- Its Campus AI Research Engineer – Deep Learning posting states that international students are encouraged to apply, that Jump accepts students eligible for CPT/OPT, and that it sponsors work visas for full-time positions (retrieved September 2026). That is more explicit than most quant firms put in a job advert, though terms vary by office and by track, so confirm with the recruiter for the specific requisition.
- What questions does Jump Trading actually ask?
- Tradermath's guide names a card-stopping game — deciding when to stop calling the next card — as a recurring Jump staple, plus coding prompts including a maximum-likelihood estimate from arrival times, ordinary least squares from scratch, a work-stealing deque, and arbitrating an A and B market data feed. Quantt adds order-book operations, cache-line awareness, lock-free data structures, adverse selection and quote adjustment.
- How long does the Jump Trading process take?
- Tradermath puts it at three to four weeks once underway; Extern says four to five rounds spanning three to six weeks; Quantt says offers typically arrive within one to two weeks of the Super Day. If a headhunter sourced you rather than a campus application, Tradermath notes it can stretch to months. Quantt says you may reapply after 12 months.