Marcus Chen, FRM
Former quantitative researcher at a Chicago-based HFT firm · 9 years in systematic futures and equities · MS Financial Engineering, University of Chicago Booth
Jump Trading is one of the most secretive and technically demanding firms in quantitative finance. Unlike more public-facing firms such as Jane Street or Citadel, Jump deliberately avoids media coverage, publishes no marketing content, and rarely makes public statements. What it does instead is quietly generate returns that place it among the most profitable proprietary trading firms in the world.
Getting an offer from Jump Trading is extraordinarily difficult — not just because the selection bar is high, but because the process itself is intentionally opaque. Candidates cannot easily find reliable prep material. Interview reports are sparse. The firm does not explain what it is looking for.
This guide synthesizes available intelligence from past candidates, quant career mentors, and first-hand accounts to give you the clearest possible picture of Jump Trading's interview process, the question types you will encounter at each stage, and how to build a preparation strategy that puts you in position to compete.
How many interview rounds does Jump Trading have?
3–4 rounds: an online assessment, one to two phone/video technical interviews, and a superday of 3–5 back-to-back interviews. Some candidates face an additional final-round call before the offer.
What does Jump Trading test in interviews?
Mental arithmetic speed, probability reasoning, brain teasers, C++/Python coding ability, and trading intuition. Jump places unusually heavy emphasis on raw computational speed — they want candidates who can do exact mental arithmetic faster than most people can type.
How long should I prepare for a Jump Trading interview?
Plan 10–16 weeks. Candidates with strong mathematics and CS backgrounds who already code fluently in C++ typically need 10 weeks if they invest heavily in mental math drills. Candidates newer to HFT-style quant interviews should budget 14–16 weeks.
Jump Trading was founded in 1999 in Chicago by William DiSomma and Paul Gurinas. The firm is privately held, discloses no financial results, and by design maintains an extremely low public profile. Jump operates across global markets including equities, futures, fixed income, options, foreign exchange, and cryptocurrency — making it one of the most broadly active proprietary trading firms in the world.
Jump is particularly known for ultra-low-latency infrastructure (co-location, custom FPGAs), systematic market-making across equities and futures, global derivatives activity on CME, ICE, and Eurex, and its proprietary technology stack built entirely in-house.
Jump's culture is extremely flat organizationally, deeply technical, and prizes individual contributors who can both research and implement. The line between "researcher" and "developer" is blurry — quantitative developers are expected to understand the trading strategies they build, and quantitative researchers are expected to write production-quality code. The firm's secrecy is not just external — internally, teams work in relatively isolated pods.
| Role | Primary Skills | Focus Area |
|---|---|---|
| Quantitative Researcher | Math/Stats/Physics PhD or BSc, Python | Signal research, strategy development |
| Quantitative Developer | C++, low-latency systems, algorithms | Execution infrastructure, trading systems |
| Quantitative Trader | Probability, mental math, market intuition | Market-making, risk management |
| Software Engineer | C++/Python, distributed systems | Core infrastructure, data pipelines |
Campus recruiting at elite universities and referrals. Resume screen prioritizes math competition background, CS fundamentals, and research experience.
Rapid-fire mental arithmetic (30–60 problems), probability brain teasers, and an algorithmic coding section. ~20–30% clearance rate.
Two to three probability questions, mental arithmetic check, one coding question. Some candidates face a second phone round before the superday.
Probability deep-dive, mental math speed round, C++ coding, trading intuition scenario, and a fit round. ~25–35% offer rate from superday.
Jump's clearest differentiator. 15–20 min of daily drills over 8–10 weeks is required — this cannot be crammed.
What is 7/13 as a decimal to 4 significant figures?
What is 127 × 43?
Simplify: (3/7) ÷ (9/14)
Computationally demanding problems that test both probabilistic setup and precise arithmetic simultaneously.
Expected rolls to see all six faces of a die (Coupon Collector)
Probability of ever reaching −1 in a biased random walk (2/3 up, 1/3 down)
Expected cards drawn until first ace in a standard 52-card deck
C++ mandatory for Quant Developer; Python for Researcher. Jump places heavy coding weight across all quantitative roles.
Implement a lock-free queue in C++
Compute rolling VWAP with O(1) updates from a real-time bid/ask stream
Monte Carlo simulator for Asian option pricing in Python
Used as a proxy for structured reasoning under unfamiliar conditions. Pattern recognition within 30–60 seconds is the bar.
12 coins, one counterfeit (heavier or lighter). Identify it in 3 balance scale weighings.
100 doors toggled — which doors are open after 100 passes?
Measure exactly 45 minutes using two non-uniform 60-minute ropes.
Scenario-based rather than structured games. Demonstrate understanding of spread economics and Bayesian price updating.
Show bid/ask for a security you believe worth $48–$52 with 80% confidence
Update your market after each trade in a sequential information game
Price a large block trade in a thinly-traded small-cap stock
Jump's culture values throughput. A quick, correct answer strictly beats a slow, elegant one. Practice solving problems fast — set hard time limits and treat going overtime as a failure mode.
You must be genuinely fluent in C++ — not 'familiar with.' Write production-quality code without reference documentation, discuss memory layout and cache performance, and reason about concurrency and thread safety.
The Coupon Collector problem, Gambler's Ruin, and the Ballot Problem appear at Jump with disproportionate frequency. Know their solutions, derivations, and generalizations. Derive Coupon Collector in under 90 seconds without writing anything down.
Jump interviewers frequently stay silent after you give an answer — they want to see if you second-guess yourself. Practice holding silence after an answer and elaborating only if asked.
Jump is a liquidity provider, not a hedge fund. They do not take long-term directional positions. Their edge is in technology speed, quantitative pricing models, and execution efficiency at microsecond timescales. Knowing this signals genuine preparation.
Most candidates spend their prep time on Xinfeng Zhou and Heard on the Street. Those books are necessary but not sufficient. Arithmetic speed is trainable only through weeks of daily drilling — it cannot be crammed. Treat it like a physical skill.
| Weeks | Focus | Resources |
|---|---|---|
| 1–3 | Mental math drilling (15–20 min/day, every day) | Myntbit mental math drills, timed flashcard apps |
| 4–6 | Probability foundations: EV, conditional probability, distributions, Coupon Collector, Gambler's Ruin | Zhou, Mosteller, Myntbit probability track |
| 7–9 | C++ algorithmic prep: STL, graphs, dynamic programming, numerical methods | Competitive programming, Myntbit quant dev track |
| 10–11 | Brain teasers + combined mental math/probability timed sessions | Heard on the Street, Myntbit timed sets |
| 12–13 | Full-length mock superdays with arithmetic under time pressure | Myntbit interview simulator |
| 14 | Rest, targeted weak-area review | Light drilling only |
Jump Trading is not looking for the candidate who has read all the right books. They are looking for the candidate who computes faster, codes more precisely, and reasons more clearly under pressure than everyone else in the pool. The interview process is designed to surface exactly those qualities — and to filter out candidates who have memorized correct answers without developing the underlying skills.
The path to a Jump offer is specific: build arithmetic speed first, then probability depth, then coding fluency, then integrate all three under timed pressure. There are no shortcuts and no ways to fake proficiency at this level.
Start with a diagnostic. Take a timed 75-minute session with mental math, probability, and coding problems to establish your baseline. Then drill daily. The arithmetic fluency and probabilistic intuition that Jump looks for are built one session at a time.
Myntbit's HFT practice set includes 650+ problems calibrated across mental math, probability, and C++ coding — each with worked solutions sourced from past candidates.