Citadel QD On-Site Coding Ladder (3 Rounds, 3-4 Problems Each)

Citadel QD onsite coding pattern reported by a senior bank quant: three coding rounds, each packing 3-4 LeetCode-style problems (mostly mediums plus 1-2 hards) into the slot. Partial credit on one hard does not block advance, but pace is the dominant grading axis. Requirements Three separate coding rounds, each ~60 minutes. Per round: 3-4 problems total. 2-3 LeetCode-medium difficulty. 1-2 LeetCode-hard. A single unsolved hard is acceptable — the reported candidate failed one hard and still progressed past this stage. The bar appears to be sustained throughput across problems rather than perfect coverage. Notes Pace is the discriminator. With 4 problems per 60-minute slot the per-problem budget is 15 minutes including reading, clarification, coding, and dry-run. Prepare to skip problems that aren't yielding within 5 minutes and circle back. Topic distribution skews toward data structures (hashmaps, heaps, trees), search / sort variations, and DP on small state spaces. Lower frequency on graph-heavy or pure-math problems. Multithreading appears in the follow-up phase, not the coding ladder — see the separate technical phone-series briefing. Coding rounds themselves stay single-threaded. Verbalize the algorithm choice before coding to bank partial credit when running out of time. The interviewer scores stated approach + working code on simple cases higher than half-finished code without a clear plan. This pattern was observed for an experienced bank quant interviewing for a Quant Dev role. NG / intern loops have a shorter ladder (1-2 coding rounds) with similar difficulty mix. Preparation Drill speed: solve 4 medium-difficulty problems in 60 minutes under time pressure, at least 3 sessions before the onsite. Calibrate the per-problem internal clock. Refresh DP, heap, hashmap, and binary-search templates so they are not the time sink. The Citadel hards lean on standard patterns dressed in domain framing rather than novel algorithms. Practice the abandon-and-return strategy: after 5 minutes with no concrete approach, write a brute-force comment and move on. Returning later with the brute-force as fallback beats spending 30 minutes on one hard. Have a 30-second "what I would do given more time" pitch ready for each problem you do not finish — it captures the partial-credit signal the interviewer needs.