Software Engineering Interviews at Trading Firms and Banks

Where the standard advice still applies

The algorithmic round is close enough to the familiar one that the usual preparation transfers. Arrays, hash maps, trees, graphs, dynamic programming, and clear complexity analysis are all tested, and the expectation that you talk while you work is the same.

One difference is worth knowing: at several trading firms the round is timed more tightly and weighted more heavily on getting to a working answer, because the trait being tested is throughput under pressure rather than elegance.

Latency is a first-class concern, not a footnote

At a consumer company, a request budget is measured in hundreds of milliseconds. On a trading system the interesting numbers are microseconds, and that changes what the interviewer wants to hear. Allocation, cache behaviour, garbage collection pauses and lock contention move from advanced topics to ordinary conversation.

You are not expected to have written a matching engine. You are expected to know why an unpredictable pause matters more than a slightly higher average, and to reason about a tail rather than a mean.

• Why p99 latency is the number that matters and an average can hide a disaster
• Where allocations happen in your language of choice, and what that costs
• Why a garbage collection pause is a different problem here than elsewhere

Correctness carries a different weight

A bug in a recommendation service degrades an experience. A bug in an order router loses money in seconds and may not be recoverable. Interviews reflect that: expect more attention to edge cases, to what happens on partial failure, and to how you would know something had gone wrong before somebody told you.

Candidates who volunteer the failure cases unprompted read very differently from candidates who wait to be asked. "What happens if this message arrives twice, or out of order" is the sort of thing worth raising yourself.

• Idempotency, and why a retried instruction is not automatically safe
• Ordering guarantees, and what breaks when they do not hold
• How you detect a fault rather than how you prevent every fault

The system design round is a different shape

The canonical exercise is to design something at consumer scale for many users. Here the axis is usually different: fewer users, far more events, much tighter timing, and a hard requirement that you can reconstruct exactly what happened afterwards.

Auditability is the requirement most candidates omit. A regulated firm must be able to say what it did, when, and why, which shapes the design in ways a photo-sharing service never has to consider.

• Market data ingestion: high volume, ordering matters, dropping is sometimes correct
• An event log you can replay, because reconstructing state is a legal requirement not a nicety
• Risk checks in the path, and the honest trade between safety and speed

The domain conversation you are not expecting

Somewhere in the process a non-engineer will ask why you want to work here rather than at a company whose product you use. It is not a formality. Firms in this space lose expensive hires to people who wanted a hard engineering problem and did not care what it was for.

You do not need to have a view on markets. You do need a better answer than compensation, and a candidate who has read anything about what the firm actually does is already unusual.

Languages, and what the choice signals

C++ dominates latency-sensitive work, Java is widespread across bank infrastructure, and Python is everywhere for research and tooling. Rust appears increasingly at newer firms. Most places let you interview in whatever you are strongest in.

What matters is depth rather than breadth. Being asked how your language manages memory, or what happens between writing a line and it executing, is common — and being fluent in one language to that depth reads better than being conversational in four.

Timelines close earlier than tech

Trading firms open early-career applications well ahead of the big tech cycle and close them on a rolling basis once the class is full. A candidate working to a technology-industry calendar can find the strongest firms finished before they start.

The tracker on this site carries live openings and stated deadlines, and trading firms are consistently among the earliest listings on it.