The Jane Street SWE Interview Guide

See how Jane Street actually runs its software engineer loop, and why it looks nothing like a LeetCode grind. The rounds hand you a plain-English spec and a few stubs, then keep changing the requirements, usually without letting you run the code. What gets scored is the interface you design, the edge cases you catch on your own, and how you hold up when the interviewer pushes back on your first idea.

• Loop length: 2 to 5 weeks
• Expected difficulty: Hard, open-ended
• Style: API design + coding
• Language: Agnostic (OCaml internally)

Jane Street levels & pay

• Software Engineer (New Grad)
• Software Engineer (Mid)
• Senior Engineer (Senior)
• Senior Engineer (Staff+)

The Jane Street SWE interview, round by round

Recruiter screen (20 to 30 min · call)

The first real conversation, usually with an in-house recruiter, and often preceded by an external headhunter call if the firm approached you first. They walk through your background, explain how the teams are organized, and are direct about the practical constraints early: the firm is in the office most days, and the role is tied to one specific city. Some tracks send a short take-home or online exercise around this point, so it is worth asking whether yours has one. Nothing technical is scored here, but the recruiter sets your schedule, and once they decide to move they move quickly.

• Ask which track you are on. Jane Street runs separate loops for software engineering, trading and research, and the question style differs so much between them that preparing for the wrong one wastes weeks.
• Relocation comes up immediately. Recruiters raise location in the first call because hiring is into a named office, and one candidate reports being asked outright what compensation a move to New York would need.
• Check whether a take-home comes first. Several tracks get an online exercise before the first live round, and candidates who did not expect one lose scheduling time discovering it.

First technical phone screen (45 to 60 min · shared editor)

One interviewer, one problem, and a plain-English specification rather than an algorithm puzzle. You are handed a few classes or function stubs and asked to build a small working piece of a system: matching buy requests against the cheapest available seller, a supermarket checkout queue, a grid game where pieces are inserted from the bottom. The interviewer then extends it two or three times, and the extensions are where the round is decided. Candidates report over and over that the code is never actually executed, so you have to walk your own edge cases by hand and out loud.

• Design the interface before the logic. One candidate summed the round up as focused on API design and clean code with only moderate algorithmic difficulty, and interviewers score the shape of that interface as closely as the logic underneath it.
• Nothing compiles and nothing runs. One debrief lists the scoring criteria outright: code cleanliness, communication, edge cases, and the ability to debug without running the code.
• The follow-ups carry the weight. A working first answer that collapses on the rollback case, the streaming variant, or the customer who cuts into another queue is the most common way candidates report failing this round.

Second technical round (60 min · call)

A second live round, usually within a week of the first. For a general software role this is another build-and-extend problem, and a performance constraint normally arrives as a follow-up rather than as the opening question. Specialized tracks branch here instead, so machine learning candidates get a modeling or optimization round, infrastructure candidates get a logs-and-latency debugging round on a mock trading system, and front end candidates get a small UI to build. A minority of candidates, almost all on trading-adjacent tracks, get a numeracy round here instead: estimation and arithmetic with no calculator.

• Performance arrives as a follow-up. The pattern is a working method first, then a nudge to make it cheap, such as re-checking only the row and column a single move could have touched instead of scanning the whole board.
• Your specialty gets tested properly. The branch round is drawn from the exact job you applied for, and candidates describe the machine learning and infrastructure versions going noticeably deeper than a generalist screen.
• Warm up arithmetic anyway. The numeracy round is standard on trading and trading-operations tracks and turns up occasionally elsewhere, and reported questions include estimating how many plastic bags are in a city.

The onsite (2 to 3 rounds · onsite)

The onsite is short by big-tech standards: two rounds in the morning and often one after lunch. Rounds run about an hour each, and candidates repeatedly describe two software engineers in the room. Each round opens with a deliberately simple problem, asks you to settle the API and the class abstractions first, then piles on follow-ups: how a ring buffer handles fragmentation, what changes once there is more than one reader or writer thread, how an arbitrage strategy stays correct at a million quote updates a second. Lunch is part of the schedule and it is also when the recruiter collects the morning feedback, so candidates who are clearly not going to pass are sometimes sent home before the afternoon.

• Abstractions first, features second. Every onsite debrief describes the same opening: an easy version of the problem, an API and class layout you agree on out loud, and only then the follow-ups that decide the round.
• Trading context is the setting, not the exam. You are not expected to know finance. The arbitrage question spells out the bids and asks in the prompt, so ask what a market feed is and then reason about it the way you would any other system with invariants.
• Coding rounds get harder here. The signature onsite problem is being handed pre-written code you cannot run and asked to change its behavior, such as making a nested-loop report O(1) while a global filter and an async callback keep invalidating your cached state.

Decision and offer (2 days to 2 weeks · after)

Feedback moves fast by industry standards. Candidates commonly hear about the next step within two to four days of a round, and final decisions after the onsite often land inside a week, though a few candidates waited two. Compensation is base salary plus a large discretionary cash bonus, with no equity of any kind, because the firm is private with no outside shareholders. If an offer comes, it is tied to the office you interviewed for rather than being open across the firm's four hiring markets.

• The turnaround is genuinely quick. Two-to-four-day gaps between rounds are the norm in candidate reports, so keep your calendar open once the loop actually starts.
• There is no stock to negotiate. The variable component is a discretionary cash bonus, so model the package as base plus bonus rather than base plus a vesting grant.
• Rejections usually arrive fast too. Several candidates report a no within a day or two of a round, though others waited one to two weeks, so silence is not a signal in either direction.

How to prepare

Build a daily probability and mental-math habit (Weeks 1 to 4)

Every day, work a handful of probability puzzles and do arithmetic and expected-value calculations without a calculator. Rotate through conditional probability, expectation, combinatorics, and simple games. Fluency here is the single biggest predictor of how the loop goes, so keep it warm the whole month.

Practice reasoning out loud (Weeks 1 to 3)

Start narrating your thinking from day one. Solve problems while speaking every assumption, dead end, and correction aloud, ideally with a partner who can throw follow-ups at you. The goal is to make your reasoning legible under pressure, because that is what interviewers actually score.

Drill language-agnostic coding and estimation (Weeks 2 to 3)

Add coding reps in whatever language you reason fastest in, prioritizing clean, correct solutions over clever tricks. Alongside them, practice Fermi estimation: decompose fuzzy questions into defensible assumptions and land a sane number, then defend it when the constraints shift.

Run full mock rounds under twists (Week 4)

In the final week, simulate the super day: back-to-back rounds mixing coding, probability, estimation, and open discussion, with a partner who keeps extending each problem. Train yourself to stay calm when the goalposts move and to keep thinking clearly across several interviews in a row.

Recently asked

Expected value of a simple dice game (reason and code)

About Jane Street

Jane Street is a proprietary trading firm that uses its own capital to trade a broad range of financial products across markets worldwide. It is privately held, with no outside shareholders and no public equity, and it has built a reputation for deep quantitative rigor, an unusual engineering culture, and a heavy reliance on the functional programming language OCaml across its entire technology stack.

The firm is known for an understated, research-driven culture that prizes clear thinking, collaboration, and intellectual honesty over hierarchy or flash. Engineering sits at the center of the business rather than in a support role, and compensation is famously high, driven by base salary plus a large discretionary bonus rather than stock. That combination of strong engineering, quantitative depth, and a distinctive in-house toolchain makes it one of the most selective places for software engineers to work.

Frequently asked

Is the Jane Street interview LeetCode?

No. The software engineering rounds are build-and-extend problems written in plain English: you are handed a few class or function stubs and asked to implement a small piece of a working system, then the interviewer keeps adding requirements. Candidates report that the code usually is not run, so clean naming, edge cases, and finding your own bugs by hand matter far more than a memorized algorithm. Probability puzzles genuinely are a Jane Street signature, but they are the spine of the trading and research loops rather than the software engineering one.

Do I need to know OCaml to interview at Jane Street?

No. Jane Street uses OCaml across its stack internally, but interviews are language-agnostic and you can code in whatever language you reason fastest in. The firm hires strong generalists and teaches its toolchain in-house, so OCaml experience is a bonus, not a requirement.

How much probability does the Jane Street software engineer interview involve?

Less than the firm's reputation suggests, if you are interviewing for a software engineering role. Probability, expected value and market-making puzzles dominate the quant trader and quant researcher loops, and they appear in the numeracy round that some trading-adjacent candidates get. Software engineering candidates far more often report ring buffer, marketplace and grid-game coding problems with no probability at all. A couple of hours of expected-value practice is reasonable insurance, but it should not be the center of your prep.

What is the Jane Street onsite like?

A comparatively short day at the office: two rounds in the morning and often one after lunch, each about an hour, frequently with two software engineers in the room. Each round starts with an easy version of a problem, expects you to design the API and class abstractions out loud, then extends it, so reported rounds include a ring buffer you can read and write strings into (what about fragmentation, what about several reader and writer threads), a file system design, and a simple exchange arbitrage strategy that then has to survive a million quote updates a second. Lunch is built into the schedule and is also when the recruiter gathers the morning feedback, so candidates who are clearly not going to pass are sometimes sent home rather than finishing the day. At least one candidate reports a virtual onsite instead of an in-person one.

How is Jane Street compensation structured?

Base salary plus a large discretionary cash bonus. Jane Street is private with no public equity, so there are no RSUs or stock options, and public salary datasets record essentially zero stock for its software engineers. Total compensation is driven by the bonus, which is why headline numbers look high next to a big tech base. Sign-on amounts are discussed on forums but are not published in any dataset that can be checked, so treat any specific sign-on figure, including the one in the ladder on this page, as unverified.

Is new-grad pay at Jane Street really that high?

It is high, and it is almost entirely cash. Entry-level software engineering packages sit at or above the top of the big tech new-grad range, but the shape is different: there is no stock, so a first-year package is base salary plus a discretionary bonus rather than a grant that accumulates over four years. Sign-on amounts are widely discussed on forums and not published anywhere reliable, so treat any specific sign-on figure you see, including the one in the ladder above, as unverified.

Does Jane Street give stock or equity?

No. As a private proprietary trading firm, Jane Street has no public equity to grant. Compensation is base plus a discretionary cash bonus, so when you model an offer, treat the bonus as the variable component rather than expecting RSUs or options.

Do I need finance knowledge to interview at Jane Street?

No prior finance or trading knowledge is required. When a question uses a trading setting, the interviewer defines it in the prompt: the reported arbitrage problem spells out what the bids and asks are before asking for a strategy. What is scored is the abstraction you pick and how you handle the follow-ups, not whether you know the instruments. Curiosity about markets helps in conversation, and asking what a term means costs you nothing.

How hard is the Jane Street interview?

Hard, but not in the way most people prepare for. The algorithmic content is moderate, and candidates repeatedly describe the problems as easier than a hard LeetCode question. What cuts people is the volume of edge cases hiding in an open-ended spec, being asked to change a design after it already works, and having to verify code that never runs. Interviewers also challenge your first idea almost immediately, so defending or abandoning it calmly is part of what is being measured.

How long does the Jane Street interview process take?

Most candidates go from the first recruiter conversation to a decision in roughly two to five weeks, which is fast for a firm with this reputation. Gaps between rounds are commonly two to four days, and several candidates report hearing about the next step the day after an interview. The slowest part is scheduling: one candidate was asked to fill in availability across a two-week window before the first live round. Decisions after the onsite usually land within a week, and rejections are often quicker than that, though some candidates waited one to two weeks for a no.

What are the mental-math questions like at Jane Street?

They live mostly in the numeracy round, which trading, trading desk operations and some quant candidates get and most software engineering candidates never see. Reported examples include estimating how many plastic bags are in a city and counting how many fours appear between one and one thousand, worked aloud with no calculator. The interviewer pushes on your assumptions as much as your arithmetic. If you are interviewing for a software role, an hour or two of practice is sensible insurance rather than a core part of your prep.

What kind of coding does Jane Street ask?

Small systems, described in plain English, with stubs already written for you. Reported problems include matching buy requests to the cheapest available seller and then rolling those trades back, a ring buffer you can read and write strings into, an infinite Connect-N board, a supermarket queue, merging two diffs in linear time, and flagging IP addresses that hit an endpoint too frequently. You pick the language. The recurring theme is that you design the interface first, the interviewer keeps extending the requirements, and you usually cannot run anything to check yourself, including on the onsite question where the class is already written and you only get to modify it.

How should I prepare for the Jane Street interview?

Practice building small systems from a written spec rather than grinding algorithm sets: pick a ring buffer, a marketplace that matches buyers to the cheapest seller, a rate limiter and a grid game, and implement each one cleanly with a real interface. Then practice changing them, because the follow-ups are the actual test, and practice modifying code someone else wrote rather than only writing your own. Work at least some sessions in a plain text editor with no compiler so you get used to finding your own bugs by reading. Add a few hours of expected-value and estimation practice as insurance, and revise your specialty closely if you applied to a machine learning, infrastructure or front end role, because the second round branches there.

What does Jane Street look for in software engineers?

Code another engineer could pick up, and a design that survives being changed. Interviewers watch how you name things, whether you handle the awkward cases without being told to, and whether you can explain a trade-off when they push back on your first idea. They also value candidates who ask questions instead of guessing at an ambiguous spec, since the specs are deliberately underdetermined. Consistency across rounds matters more than one standout answer.

Does Jane Street have an online assessment?

Sometimes, and it depends on the track. There is no timed, auto-graded LeetCode-style test for software engineering roles, but candidates on several tracks report a take-home or online exercise before the first live round, including a spreadsheet-based assessment for trading desk operations that arrived the day after the resume went in. At least one software engineering candidate describes passing an online test and being invited to the onsite off the back of it. Ask your recruiter directly, because whether you get one is the biggest scheduling variable early in the process.

Where does Jane Street hire software engineers?

Four offices: New York, which is the headquarters and by far the largest, then Hong Kong, London and Singapore. All four carry live Software Engineer postings on Jane Street's own job board, at new grad and experienced level. The firm also has offices in Amsterdam and Chicago, but neither currently lists any role of any kind on that board, so plan on one of the four. The interview style is the same across offices, and compensation is set to the local market.

Jane Street interview questions on Karavine