← Back to blog HRT VO Onsite Complete Guide｜Hudson River Trading Five-Round Algo Dev + Brainteaser + System Design
VO HRT

HRT VO Onsite Complete Guide｜Hudson River Trading Five-Round Algo Dev + Brainteaser + System Design

2026-05-16

Hudson River Trading (HRT) is a top-tier quant prop trading firm based in NYC, in the same league as Citadel, Jane Street, and Two Sigma. OA pass rate is roughly 8-12%, but the onsite is the real meat grinder — five rounds, any of which can knock you out. This article focuses on the post-OA onsite flow for HRT's three engineering-track Algorithm Developer (Algo Dev) / Quant System / Core Infrastructure roles.

If you haven't passed the HRT OA yet, start with our HRT OA 2026 guide.

HRT Onsite Pipeline Overview

Round Duration Content Pass Rate
1. Tech Phone Screen 60 min Coding + 1 brainteaser 100% → 50%
2. Onsite #1: Algo + Coding 75 min Mid-LC algorithm + perf discussion 50% → 30%
3. Onsite #2: System Design 75 min Low-latency C++ system + concurrency 30% → 18%
4. Onsite #3: Brainteaser + Probability 60 min 4-6 probability puzzles, mental math 18% → 12%
5. Onsite #4: Behavioral + Culture Fit 60 min Project deep-dive + team fit + reverse Qs 12% → 8%

HRT specifics:

Onsite #1 — Algo + Coding (Looks Like LC, but Pitfalls Are in Performance)

Sample Question

Given an integer array prices of length n (each entry a tick price), compute the sum across all length-k windows of "price range" (max − min). Constraints: n ≤ 1e7, k ≤ 1000.

Two Layers of Difficulty

Layer 1 (30-point pass): two monotonic deques for max and min.

#include <deque>
#include <vector>
using namespace std;

long long range_sum(const vector<int>& prices, int k) {
    deque<int> dec, inc;  // decreasing for max, increasing for min
    long long total = 0;
    for (int i = 0; i < (int)prices.size(); ++i) {
        while (!dec.empty() && prices[dec.back()] <= prices[i]) dec.pop_back();
        dec.push_back(i);
        while (!inc.empty() && prices[inc.back()] >= prices[i]) inc.pop_back();
        inc.push_back(i);
        if (dec.front() <= i - k) dec.pop_front();
        if (inc.front() <= i - k) inc.pop_front();
        if (i >= k - 1) total += prices[dec.front()] - prices[inc.front()];
    }
    return total;
}

Layer 2 (what HRT actually grades): the interviewer follows up:

Difference between HRT Algo Dev and a normal SDE interview: writing the O(n) algorithm is just the entry. You need to discuss cache lines / branch prediction / SIMD / lock-free patterns to clear the bar.

Prep Path

Onsite #2 — System Design (HRT's Signature Round)

Sample Question

Design a market data ingestion pipeline. Input: multicast UDP from NYSE / NASDAQ (~5M packets/sec). Output: normalized OrderBook state pushed to 50 downstream trading-strategy processes. SLA: 99.9th percentile latency < 30 microseconds.

Preferred HRT Answer Structure

1. Workload — 5M pps × 200 bytes ≈ 1 GB/s, NUMA-aware
2. Ingress — kernel bypass (DPDK / Solarflare); single-core single-queue RSS
3. Parse — preallocated ring buffer; zero-copy parser; FBS / FlatBuffers
4. State — per-symbol L2 OrderBook; hot data cache-line aligned
5. Distribution — shared memory + spinlock (or lock-free SPMC queue)
6. Observability — CPU pinning + perf counters + tail-latency histogram

Must-discuss topics:

Penalty answer: jumping straight into Kafka / RabbitMQ. HRT engineers immediately mark "not aware of low-latency constraints."

Onsite #3 — Brainteaser + Probability (All Mental Math, Easiest to Fail)

Common Question Types

Q1: 100 pirates rank from 1 to 100 distribute gold by lowest-rank-first proposal, majority vote required. Find the maximum gold the captain can secure.

Q2: Run 1024 trials of 10 fair coin flips. For each trial record the longest run L of consecutive heads. Find E[max L].

Q3: Sample uniformly from [0, 1] until cumulative sum > 1. What's the expected number of samples?

Strategy

Dimension Wrong Right
Time Think silently for 5 min State a hypothesis within 30 seconds, then derive aloud
Tool Mental-only Use the whiteboard / shared screen; HRT Zoom provides one
Pace 15 min on one question Avg 8-10 min/question; must ask for a hint after 5 min stuck
Confidence Stay silent if unsure State your best guess, then "what I might be missing..."

HRT brainteasers are not about being right, they're about your thinking process — a wrong answer with clear reasoning often outscores a right answer with muddled derivation.

Recommended Books

Onsite #4 — Behavioral + Culture Fit

What HRT Cares About

Common Questions and Approach

Question Recommended Angle
"Why HRT over Citadel / Jane Street?" Cite specific tech differences (HRT = C++ infra; Citadel = multi-strategy; JS = OCaml + research)
"Your most challenging debugging story?" Pick one with performance / concurrency / numerical precision — the lower-level, the better
"When did you change your mind?" Must-prep — HRT scores intellectual humility
"Any questions for me?" Have 3: 1 about onboarding cadence, 1 about their personal work, 1 about a specific tech decision

6-Week Prep Path (Assuming OA Cleared)

Week Focus Resources
W1-2 Algo + perf 30 LC Hard + Agner Fog manual + 3 CppCon talks
W3 System Design "Designing Data-Intensive Apps" ch. 5, 7 + 5 HRT blog posts
W4 Brainteaser Full Crack + green book ch. 5-9
W5 Mock 2 full sets of 4-round mocks
W6 BQ + reverse Qs 6 STAR stories + 3 reverse questions

FAQ

Q1: What's the actual HRT VO pass rate?

Onsite pass rate after clearing OA is ~30-40% — not actually low — but OA itself eliminates 88-92%. Combined, resume-to-offer conversion is roughly 1.5-3%, on par with Jane Street and Citadel SDE.

Q2: Do I have to use C++ for HRT VO?

Algo Dev / Core Infrastructure: strongly recommended (team primary language). Other languages are tolerated but you'll need a migration plan in BQ. Quant System track accepts Python (many ML / strategy engineers are Python-only). Practical advice: C++ for Algo + System Design rounds; Brainteaser / BQ rounds don't involve code.

Q3: Does HRT sponsor H1B?

Yes. HRT is among the more sponsorship-friendly quants — roughly 60% of engineers are non-US-born. However, HRT's onsite cycle (June) misses the main H1B lottery, so new hires typically bridge with OPT + STEM extension.

Q4: What are HRT salaries?

2026 NYC base + bonus (engineering tracks, including sign-on amortization):

vs FAANG: roughly on par with Jane Street / Citadel engineering, 50-100% higher than FAANG L4-L5.

Q5: How long is the cooldown after a HRT VO failure?

12 months for the same track. Switching tracks doesn't bypass it — HRT shares scoring internally. A strong rejection on a specific round (e.g., brainteaser) typically extends the wait to 18 months.

Q6: Is the HRT work environment really "hellmode"?

Not hell, but tougher pace than FAANG. Typical week: 50-55 hours at the NYC office; peak weeks reach 65 hours. Pay covers it well — Glassdoor culture 4.4, Blind recommend rate 80%+, far above most trading firms. Manager is the key variable: good managers run barely-OT teams; bad managers run heavy-OT ones.

Contact

If you're prepping HRT, Citadel, Jane Street, Two Sigma — top-tier quant engineering — the OA is just the first gate; system design + brainteaser onsite is the actual filter. We've curated 2025-2026 HRT engineering Onsite bank: 50+ real questions + system design templates + 60-question brainteaser solutions + mock services.

Add WeChat Coding0201, get the HRT Onsite bank and book mocks.