Jane Street Interview Guide (2026)

1Snapshot, read this first#

The one thing to internalize about this firm:

• They are not hiring you for what you already know. They are hiring you for how you think when you do not know.
• Every round is secretly a test of that.
• A trader who only quotes confidently when the answer is obvious is worthless. The money is made in the gray.
• So the interview is deliberately built to push you into the gray and watch what you do there.

2Getting the call (the part most guides skip)#

Jane Street hires overwhelmingly from its own channels. Resumes get read carefully, but the front door is their own pipeline.

• Most new grads come in through campus recruiting and the fall/winter internship pipeline.
• The internship-to-full-time conversion rate is high, so the internship is the real front door.
• Events like their puzzle hunts, Estimathon, and the Electronic Trading Challenge are where the real openings are.
• A strong showing at an onsite event or a clean Estimathon round gets a recruiter emailing you, which skips the resume lottery entirely.
• If you are a student, going to those is not a nice-to-have. It is the single best thing you can do.

Referrals work differently here than at big tech.

• There is no referral bonus culture.
• What helps: a current employee who genuinely knows your work and will vouch.
• A cold LinkedIn referral from a stranger does almost nothing.
• A referral from someone who cannot speak to anything you built can hurt, because the recruiter's first question to the referrer is "how do you know them."

On the resume, they care about depth and signal, not buzzwords. One real project you understand end to end beats ten half-finished ones. Here is the shape of a bullet that gets a second look:

Built a lock-free order-book in C++ handling 2M updates/sec, cut p99 match latency from 40us to 9us by replacing the std::map level structure with a flat array keyed on price ticks.

Why that bullet works:

• Numbers, a real bottleneck, a specific decision, a measured result. That is the whole formula.
• An interviewer can open it up: why was std::map slow here, what did the flat array cost you in memory, how did you measure p99.
• If you wrote it, you answer all three in your sleep. If you padded it, you stall on the second one, and they feel it.

Keywords and signals that actually land:

• Put the obvious keywords in if they are true: C++, OCaml (yes, really, they are famously an OCaml shop), Python, data structures, algorithms, probability, systems. Do not stuff them.
• If you have competitive programming ratings (Codeforces, USACO, ICPC, Putnam), put them near the top.
• A Codeforces rating above 1900 or an ICPC regional finish is worth more than a FAANG internship on this particular resume, because it maps directly to the speed and correctness the loop measures.

3The loop at a glance#

How the tracks differ:

• Trader-track loops lean more on probability, mental math, and the market-making game.
• SWE-track loops lean more on coding and a bit of systems.
• The undecided "SWE or Trader" track gets a mix and they figure out where you fit.

4The coding or technical rounds#

The coding here is closer to competitive programming than to LeetCode pattern-matching.

• Expect clean algorithmic problems where the trick is seeing the right data structure, not reciting a template.
• A typical problem reads like a puzzle, not a tagged entry.
• You get a scenario and have to spot that it reduces to, say, a monotonic stack or a priority-queue sweep, with no tag telling you which bucket it lives in.

Topics they reuse:

• Arrays and two-pointer, hashing, recursion and backtracking.
• Dynamic programming (real DP, not toy cases).
• Graph traversal (BFS/DFS, shortest path), intervals, heaps and priority queues.
• Anything touching complexity tradeoffs.
• For C++ SWE candidates, go deeper: cache behavior, memory layout, why a vector beats a linked list in practice, maybe a lock-free or concurrency question.

A real one I have asked: "You have a vector of a million structs and you iterate summing one field. Now the struct grows from 16 bytes to 200 bytes. What happens to your runtime and why?" The answer is cache lines, and it should come out fast.

How to practice:

• Do timed sets on Codeforces (Div 2 A through D is a great range) and the harder half of a LeetCode topic list.
• Practice with a timer and practice talking the whole time, even alone.
• That feels silly. Do it anyway.
• The skill is not "solve the problem," it is "solve the problem while narrating." Those are different skills that have to be trained together.

How to win the round:

• Restate and confirm constraints first. "So n can be up to 10^5, values fit in 32 bits, is that right?" This also tells you the intended complexity: n up to 10^5 means roughly n log n, not n squared.
• State your approach and its complexity out loud before writing. Let them redirect you early. A good interviewer nudges you off a dead end in the first two minutes if you give them the chance by talking.
• Write it clean the first time. Good names, no dead code. A variable named i in a tight loop is fine; a sorting comparator named thing2 is not.
• Test it yourself. Walk a small input by hand, check empty, single element, duplicates. Finding your own bug is a strong positive signal. In a trading firm, catching your own mistake before it ships is the whole job.
• When stuck, narrate what you are checking. "When I run this loop, what I am really verifying is that the invariant holds after each step" tells the interviewer you understand your own code.

Probability and brain-teasers#

This is where Jane Street separates itself, and where most candidates quietly fail. Expect expected value, conditional probability, linearity of expectation, simple Markov-chain reasoning, fair-value of a game, and classic puzzles in new clothes.

A flavor of what shows up:

• You roll a die repeatedly and sum until the sum exceeds 100. What is the expected final value over 100?
• Two players alternate flipping a coin; first to get heads wins. What is the probability the first player wins?
• Expected number of rolls to see all six faces of a die (coupon collector).
• A stick breaks at two uniform random points. Probability the three pieces form a triangle.

Do not memorize answers. Memorize methods.

• They will change a number and watch you fall apart if you only memorized.
• Keep the tools: linearity of expectation, states and transitions, symmetry arguments, conditioning on the first step.
• Say your reasoning out loud and update it when they push.
• "Let me define the state as the current sum, then condition on the next roll" is the kind of sentence that earns the round.

Here is the coin-flip one worked the way they want to hear it:

• Let p be the probability the first player wins.
• On the first flip, the first player gets heads with probability 1/2 and wins immediately.
• Otherwise, with probability 1/2, the game is now the same game with the second player in the first-player seat, so the first player now wins with probability 1 minus p.
• That gives p equals 1/2 plus 1/2 times (1 minus p), which solves to p equals 2/3.
• What earned the round: I did not recall "2/3," I set up a one-step recursion out loud. Change the coin to land heads with probability q and I redo it on the spot, because I kept the method and not the number.

5The market-making game and CS fundamentals#

For trader and undecided tracks, the signature round is a market-making game.

• The interviewer runs a game with hidden information.
• Classic version: there is a quantity you cannot see (sum of cards in a deck, population of a country, dice under a cup).
• You quote a bid and an ask. They trade against you. You update your quotes as information arrives.

What they are watching:

• Do you make a reasonable market, or freeze?
• Do you update correctly when someone hits your bid (that is information, it means your price was probably too high)?
• Do you manage risk, or keep quoting after you are badly long or short?
• Do you stay calm and keep reasoning when it goes against you?

A simple framework:

• Quote around your best estimate of fair value.
• Set your spread to your uncertainty (wider when you know less).
• Move your quotes immediately when a trade tells you something.
• If they keep lifting your offer, you are too cheap, raise it. Talk through every adjustment.

Walk it concretely. Say the hidden number is the sum of two dice, so fair value is 7 and the range is 2 to 12.

• With no information I quote 6 at 8, a two-wide market around fair value.
• If the interviewer buys at my 8, they think it is worth more than 8, so I slide up and quote 7 at 9.
• If they then sell to me at 7, I am now long and the signal flipped, so I widen and lower, maybe 5 at 7, to protect against being wrong and stop bleeding.
• Every move has a spoken reason: "you lifted my offer, so I am raising both sides and keeping the spread wide because I am still uncertain."
• The candidates who pass sound like a running commentary. The ones who fail go quiet and let their position drift.

A quote with no spoken reason behind it is just a guess wearing a number.

For the SWE side, you may also get CS fundamentals mixed in. Keep answers concrete, not textbook.

• How a hash map resizes, how the OS schedules threads, what a cache miss costs you, TCP vs UDP and why latency-sensitive systems care.
• "A cache miss is roughly 100ns, a main-memory fetch, which is why we keep the hot path in contiguous memory" beats a definition.
• On networking: "UDP skips the handshake and retransmit, so it loses packets but never head-of-line blocks, which is why exchanges feed market data over UDP multicast and keep TCP for order entry where you must not drop."

6Behavioral and culture fit#

The fit round is real, not a formality. Their values in practice:

• Intellectual honesty above all.
• Curiosity.
• Collaboration over heroics.
• A flat culture where a first-year can and should argue with a senior if the senior is wrong.
• They dislike ego, and dislike bluffing even more.

The single most important thing in every round: when you do not know, say so, then reason toward it.

• "I am not sure, here is how I would figure it out" scores higher than a confident wrong answer.
• Bluffing is close to disqualifying here, because trading real money on a bluff is how firms blow up.
• A trader who says "I was 80 percent sure" when they were actually guessing costs the firm millions.
• So the interview treats a confident wrong answer as a serious red flag, not an honest miss.

Prepare three or four STAR stories:

• A hard technical problem.
• A disagreement you handled well.
• A time you were wrong and caught it (this is the one they care about most, so skip the fake-humble "I worked too hard").
• A time you taught yourself something hard.
• Keep a number in the result. On the "wrong" story, pick a real mistake: say what it cost, how you caught it, what you changed.

A worked STAR example:

• Situation: Our service was dropping messages under load during market-open spikes, about 3 percent loss at peak.
• Task: I owned reliability for that pipeline and had to stop the loss without adding meaningful latency.
• Action: I profiled it and found the consumer blocked on synchronous disk writes. I moved to a bounded in-memory ring buffer with a background flush thread, and added backpressure so producers slowed instead of dropping.
• Result: Message loss went from 3 percent to zero across the next three market-opens, and p99 latency actually dropped 15 percent because we stopped blocking the hot path.

Why that one lands: short, specific, honest about the tradeoff, ends on a measured result. The action names a real decision (ring buffer plus backpressure) and the result carries two numbers, one of which was a pleasant surprise. Volunteer that surprise. It shows you measured instead of assumed.

7Your prep plan#

4-week plan:

• Week 1: Relearn core DSA. One topic a day, 4 to 6 problems each, timed, talking out loud.
• Week 2: Probability and expected value. A problem set daily. Drill the methods, not answers.
• Week 3: Mental math (no calculator, fast 2-digit multiplication and percentages) and market-making games with a friend or by self-simulation.
• Week 4: Full mock loops. Mix one coding, one probability, one estimation per sitting. Fix your weakest area.

2-week plan:

• Days 1 to 5: DSA mediums/hards, timed, out loud.
• Days 6 to 9: Probability methods plus daily mental math drills.
• Days 10 to 12: Market-making game practice and 2 mock interviews.
• Days 13 to 14: Review mistakes, prep STAR stories, rest before the loop.

2-day crash plan:

• Day 1: 8 to 10 classic probability problems (coupon collector, coin-flip games, expected value), 5 DSA mediums, and 30 minutes of mental math.
• Day 2: One market-making game walkthrough, 3 DSA problems out loud, write your STAR stories, sleep early.

8Why people get rejected at Jane Street#

• Bluffing a probability answer instead of admitting uncertainty and reasoning. This is the number one killer.
• Coding in silence. If I cannot follow your thinking, I cannot pass you, even if the code works.
• Sloppy mental math. Slow or wrong on basic arithmetic reads as a real problem for a trading role.
• Memorized puzzle answers. They change the numbers and you have nothing underneath.
• Not updating in the market-making game. Someone hits your bid three times and you keep the same quote. That shows no feel for information.
• Ego. Getting defensive when corrected, or talking down to the interviewer. It is a flat, collaborative place and they screen hard for it.

Read that list again and notice the pattern:

• Only one of the six is about raw technical skill (sloppy math).
• The other five are about how you behave under uncertainty and correction.
• That is the real bar. You can be the strongest coder in the pool and still fail by going quiet or getting defensive.

9The offer, comp and negotiation#

Jane Street pays at or above the very top of the market, and a large chunk is a discretionary bonus, so year-one numbers vary a lot.

• US new grad SWE: commonly 200k to 450k+ total in year one with signing and bonus.
• US trader: first-year total frequently 300k to 500k+, with the bonus being the main variable.
• London: roughly 125k to 200k+ GBP total for new grads, scaling fast with performance.
• APAC (Hong Kong) and India: smaller presence, APAC hiring runs mostly through Hong Kong; packages are strong for the local market but there is less public data, so ask directly.

Why the range is so wide, especially for traders:

• The bonus is genuinely discretionary and tied to performance and firm results.
• There are no public RSU grants or vesting schedules like at big tech, because Jane Street is private.
• Two people hired into the same seat can end the year far apart.
• Do not anchor on the top of the range you saw on a forum. Anchor on the base and the signing, the parts actually promised in writing.

Negotiation tips specific to Jane Street:

• They lead with strong numbers, so there is less room than at a company that lowballs. Competing offers from other top quant firms (Citadel, HRT, Optiver, Two Sigma) are the lever that actually moves them.
• Focus on the signing bonus and start date flexibility rather than fighting the base. Those are more movable for new grads.
• Be honest and specific about competing offers. They respect directness and will often match or beat a real one.
• Do not bluff a number you do not have. These people price uncertainty for a living. Invent a competing offer and they may ask for the letter, and now your candidacy smells like the exact thing they screened against all day.

10FAQ#

Do I need to know OCaml? No, not to get hired. Jane Street is famously an OCaml shop and SWEs learn it on the job, but interviews let you use a language you are comfortable with (usually Python or C++). Functional programming concepts help you ramp up and read as a nice signal, but they are not a gate. If you want a cheap edge, spend a weekend writing a small OCaml program so you can talk about pattern matching and immutability with a straight face. It will not win the loop, but it reads as genuine curiosity, which they like.

Is it harder to get in as a trader or a SWE? Both are very selective. Trader loops weight probability, mental math, and the market-making game more heavily; SWE loops weight coding and systems. Many new grads apply to the undecided track and the firm places them. If you are unsure which you are, the undecided track is usually the smart call, because it lets the strongest signal you give decide your fit instead of forcing you to self-sort early.

How much competitive programming do I need? It helps a lot but is not required. The coding bar is medium-to-hard and fast. If you have never done timed contests, do a few weeks of Codeforces before the loop; the speed and problem-solving transfer directly. Getting comfortable finishing Div 2 A through C under the clock is enough to walk into the coding rounds without the timer rattling you.

Do they hire new grads with no finance background? Yes, actively. They assume no finance knowledge and teach it. They care about raw problem-solving, honesty, and curiosity far more than whether you know what a put option is. Do not waste prep time memorizing options theory. Spend it on probability and mental math, which is what they actually test.

What is the one skill I cannot fake? Mental math and intellectual honesty. You can cram DSA. You cannot fake fast, accurate arithmetic under pressure, and you cannot fake the instinct to say "I do not know, here is how I would find out." Build both now.

Want a plan dated to your actual interview day, built around your weakest area? Map out your weeks backward from the loop and start with the round that scares you most.