How Jump Trading Interviews in 2026: Process, Questions & What They Score

1. 01

   Write Python that computes the maximum likelihood estimate from a series of event arrival times.

   Tests deriving a likelihood and coding its closed-form maximizer, such as the mean inter-arrival time for a Poisson rate.

2. 02

   Implement ordinary least squares regression from scratch in Python.

   Tests solving the normal equations or a QR decomposition in NumPy, with attention to conditioning, without a library fit.

3. 03

   Given signed integer values and a window size, return the mean of every full window, updating in constant time per step.

   Tests a sliding-window running sum with constant-time updates over signed integers.

4. 04

   In a live session, take a time-series dataset and build a predictive model end to end in pandas, then debug and complete a long pre-written ML script.

   Tests pandas feature engineering with rolling windows, leak-free time-series splits and regularized regression.

Reported problems, listed so you know what to expect. Practice them in your own editor or on the platform you prefer; SupaCV's practice mode coaches how you talk through them.

1. 01

   Split an array into contiguous parts so that the variance (or the maximum subarray variance) is minimized.

   Tests O(n) running variance updates and a k-group dynamic program for optimal contiguous partitions.

2. 02

   Build a stack-based calculator that evaluates Reverse Polish Notation, supporting push, pop and DUP operations and handling error cases such as overflow.

   Tests stack-based expression evaluation, DUP and POP semantics, and error handling for overflow and bad input.

3. 03

   Parse the output of ls -l (Unix permission strings) and return the names of files that are read-only.

   Tests careful string parsing of fixed-format Unix permission records and edge cases.

4. 04

   Given numerators and denominators in two arrays, find the most frequent fraction value (the largest group of equal fractions), handling signs, zero and overflow.

   Tests reducing fractions with gcd and sign normalization, then hashing to find the largest equal group.

5. 05

   Find the largest subset of integers whose bitwise AND is nonzero.

   Tests per-bit counting: the best subset shares one set bit, so count integers by bit position.

6. 06

   Convert a series of buy and sell indicator signals into a position series.

   Tests a single-pass simulation that turns buy and sell signals into a running position.

7. 07

   Knapsack-style stock picking: with a fixed budget, choose which stocks to buy to maximize total value.

   Tests 0/1 knapsack dynamic programming with a budget as the capacity.

8. 08

   Design an order book that supports add, execute and cancel(order_id) efficiently, with cancel in O(log N).

   Tests an id-to-node hash map plus sorted price levels so cancel costs O(log N).

9. 09

   Arbitrate the A and B feeds: two redundant market-data feeds carry the same messages, and you merge them into one clean stream.

   Tests merging redundant sequenced streams with deduplication by sequence number and gap handling.

Reported problems, listed so you know what to expect. Practice them in your own editor or on the platform you prefer; SupaCV's practice mode coaches how you talk through them.

1. 01

   Cards are dealt one at a time from a shuffled deck and you may stop at any point; you win if the next card is red. What is your best stopping strategy and your probability of winning?

   Tests optional stopping on a shrinking deck: the next-card-red chance is a martingale, so no stopping rule beats one half.

2. 02

   Of 100 coins, 99 are fair and one has two heads. You pick one at random and flip it 10 times, getting 10 heads. What is the probability you picked the two-headed coin?

   Tests Bayes' rule: a 1 percent prior updated by a likelihood ratio of 2 to the 10th power after 10 heads.

3. 03

   You may pay a fee for each draw from a known distribution and stop whenever you like. What stopping rule maximizes your expected payoff, and what is a fair price for the game?

   Tests backward induction for an optimal stopping threshold and valuing the game as its expected payoff net of draw fees.

4. 04

   In a Markov problem where balls of different colors are repainted until all of the balls match (a three-color version is also listed), what is the expected number of steps?

   Tests modeling color counts as a Markov chain and solving expected time to absorption with first-step analysis.

5. 05

   In the guess-one-third-of-the-mean game, where the winner is the player closest to one third of the average of all guesses, what should you guess? A variant has a mixed strategy over the values 0, 1 and 100.

   Tests iterated reasoning in a p-beauty-contest game, including a mixed strategy when guesses are limited to a few values.

6. 06

   Flip a fair coin repeatedly. Which pattern is more likely to appear first, HTH or HHT?

   Tests Markov chain or martingale analysis of competing coin patterns, in the style of Penney's game.

7. 07

   Gambler's ruin on a fair coin: starting with a given stake, what is the probability of reaching a target before going broke?

   Tests random-walk absorption: ruin probability from a martingale or difference equation on a fair coin.

Reported problems, listed so you know what to expect. Practice them in your own editor or on the platform you prefer; SupaCV's practice mode coaches how you talk through them.

1. 01

   What range of values can a common pairwise correlation take among n variables?

   Tests positive semi-definiteness of an equicorrelation matrix, which bounds the common correlation below by minus 1/(n-1).

2. 02

   What is the inverse of an equicorrelation matrix?

   Tests exploiting identity-plus-rank-one structure, via Sherman-Morrison or eigenvalues, to invert in closed form.

3. 03

   A covariance matrix estimate is nearly but not quite positive semi-definite. How do you fix it?

   Tests eigenvalue clipping or nearest-PSD projection to repair a noisy covariance estimate.

Reported problems, listed so you know what to expect. Practice them in your own editor or on the platform you prefer; SupaCV's practice mode coaches how you talk through them.

1. 01

   A market maker quotes a tight bid-ask spread and a flurry of one-sided buys hits the offer. How do you adjust your market?

   Tests adverse selection: widen the spread and skew quotes up when one-sided flow may signal informed buying.

Reported problems, listed so you know what to expect. Practice them in your own editor or on the platform you prefer; SupaCV's practice mode coaches how you talk through them.

1. 01

   Implement a simplified std::vector in C++: amortized O(1) push_back or emplace_back growth, the rule of five, reserve, and exception safety for element types that are move-only or throw on copy.

   Tests dynamic array internals: amortized growth, raw storage with placement new, rule of five and exception safety.

2. 02

   Explain the amortized and worst-case time complexity of vector push_back, how unordered_map behaves, and how virtual functions are dispatched.

   Tests C++ container complexity, amortized versus worst-case cost, and vtable-based virtual dispatch.

3. 03

   Implement a lock-free single-producer single-consumer queue (or ring buffer) in C++ and explain how you avoid false sharing.

   Tests atomics with acquire-release ordering, a ring buffer and cache-line padding against false sharing.

4. 04

   Implement a work-stealing deque in C++.

   Tests a concurrent deque with owner-side push and pop, thief-side steal and careful atomic ordering.

5. 05

   Implement malloc, a bucketed allocator or a fixed-size memory pool.

   Tests allocator design with size-class buckets or fixed pools, free lists, alignment and fragmentation trade-offs.

6. 06

   You have a struct with three 4-byte integers and a 1-byte flag, and the hot path reads only the first integer and the flag. How would you lay out the struct, and why?

   Tests cache locality: group hot fields together and keep a frequently written flag off shared cache lines.

7. 07

   Walk through what happens when a market-data packet arrives at the network card, through to the trading code in a kernel-bypass setup.

   Tests kernel-bypass knowledge: NIC to user-space rings, polling instead of interrupts and zero-copy parsing.

8. 08

   What does memory_order_acquire actually guarantee on x86, ARM and RISC-V?

   Tests memory-model knowledge: acquire semantics and how x86 strong ordering differs from ARM and RISC-V.

9. 09

   Design a system that processes one million messages per second with microsecond latency.

   Tests latency-budget design: lock-free queues, core pinning, no hot-path allocation and batching.

Reported problems, listed so you know what to expect. Practice them in your own editor or on the platform you prefer; SupaCV's practice mode coaches how you talk through them.

1. 01

   Why Jump specifically, rather than another HFT firm such as Hudson River Trading or Citadel Securities? (For software candidates, also: why C++?)

2. 02

   Walk me through your resume and one past project or research piece in depth.

3. 03

   Tell me about a time you were under significant time pressure and made a mistake.

4. 04

   Tell me about a time a system you owned failed in production.