10 Jane Street Software Engineer (New Grad) Interview Questions (2026)

Outline

STAR. Pick a real bug with non-obvious root cause. Walk through: how you reproduced, narrowed down (binary search through code, log inspection, profiling), what the actual cause was, how you verified the fix. Jane Street values methodical debugging — heroic guesses score lower than structured investigation.

Outline

Specific reasons: OCaml and functional programming culture, tight engineer-trader collaboration, immediate feedback loop (trades reveal whether your code works), no clients, small mid-sized firm rather than tech-giant scale. Avoid 'high pay' as the lead. Reference Jane Street's open-source contributions (Base, Core, Async).

Outline

Track seen elements in a set and running sum. For each element: if not in set, add to set and add to sum; append current sum to result. O(n) time, O(n) space. Walk through example. In OCaml: use a Set module and fold over the list. Mention immutability tradeoff if asked.

Outline

Recursion. At each node, compute height of left and right subtrees. Diameter through this node = left_height + right_height. Track global max. Return height to parent. O(n) time, O(h) space. Walk through example. Note: diameter is in edges, not nodes — clarify with interviewer.

Outline

Expand-around-center: for each index, expand outward for odd-length and even-length palindromes. Track longest. O(n^2) time, O(1) space. Mention Manacher's O(n) algorithm but note it's rarely required in interviews — interviewer just wants to see clean O(n^2).

Outline

Referential transparency: an expression can be replaced by its value without changing program behavior. Pure functions (no side effects) are referentially transparent. Matters for parallelism: pure functions can be parallelized trivially — no shared state, no race conditions, results are deterministic. OCaml/Haskell encourage this style. Discuss why imperative code is harder to parallelize.

Outline

Closure: function bundled with its lexical environment (captured variables from enclosing scope). Allows functions to retain access to variables outside their immediate scope. Example: a function returning another function that increments a counter — the counter is captured. Discuss in OCaml/JS/Python; mention how closures enable currying and partial application. Memory implications: captured variables stay alive as long as the closure does.

Outline

Sum is even iff both dice have the same parity. P(both even) = (1/2)(1/2) = 1/4. P(both odd) = 1/4. Total = 1/2. State the parity argument cleanly. Be ready for follow-ups: weighted die, more than two dice, sum divisible by 3.

Outline

Mutex around the underlying queue. Producer locks, enqueues, signals condition variable, unlocks. Consumer locks, waits on condition if empty, dequeues, unlocks. Walk through with pseudocode. Discuss lock-free alternative using atomic operations and CAS — much harder to get right, especially around ABA problem. For correctness in interviews, mutex+CV is the safe choice.

Outline

TCP: connection-oriented, reliable, ordered, flow control, congestion control. UDP: connectionless, unreliable, no order guarantees, low overhead. Trade-off: TCP for correctness-critical (HTTP, SSH), UDP for latency-critical (DNS, video, trading market data). Discuss why high-frequency trading often prefers UDP — multicast efficiency and lower latency. Mention TCP's slow-start as a real concern in latency-sensitive systems.