Algorithm Questions That Appear in Quant Developer Interviews

Quant developer interviews recast standard algorithms as market data problems, and the follow-up question is where the interview is actually scored.

Quant developer interviews draw from the same catalogue of arrays, trees, graphs and dynamic programming that any software engineering interview does, but every problem arrives dressed in market data rather than presented as an abstract exercise. A candidate who has drilled two hundred generic problems still has to notice that a merge of sorted lists is a merge of price ticks from four exchange feeds, and that a binary search is a search for the best resting price in a sorted array of quotes.

The framing is not decoration. An interviewer at a firm such as Jane Street or Optiver is checking whether a candidate can translate a business scenario into the correct abstract structure without being told which chapter of the textbook it belongs to. In a general software interview the category is often given away by the phrasing of the question. In a quant developer interview the phrasing hides the category on purpose, because a production trading system never announces which data structure solves it either.

The three problems below appear in some form at most firms that hire quant developers, each with a standard approach, a standard complexity, and a standard follow-up that separates a candidate who memorized a solution from one who understands why it works. Practicing them as a family, through timed sets on interview-questions and repetition on practice, builds the habit of spotting the underlying structure fast.

State the problem precisely: a system receives k separate streams of timestamped price ticks, one per venue, each already sorted by time internally, and must produce a single globally sorted stream. The standard approach keeps a min-heap of size k, one entry per stream holding its current head tick, popping the earliest tick and pushing the next tick from that same stream. This runs in O(n log k) for n total ticks, which beats a plain sort at O(n log n) once k is much smaller than n, as it usually is with four or five venues feeding one book.

The standard follow-up asks what happens when a single feed can deliver a tick out of order because of jitter on that venue's line. A heap alone assumes every stream is internally sorted, so the interviewer is looking for a bounded reorder buffer per stream and a watermark that only releases a tick once every stream has advanced past its timestamp, trading a small fixed delay for a guarantee of order.