# RBC Capital Markets Interview Guide (2026): Software & Strat Roles

Source: https://www.techinterview.org/companies/rbc-capital-markets/
Updated: 2026-07-03 · techinterview.org

Start with the one fact that tells you what kind of shop this is: RBC Capital Markets runs an electronic execution platform called Aiden that uses deep reinforcement learning to work equity orders, and it was built in-house by Borealis AI, the bank's own machine-learning research institute. Most bank tech orgs can't point to anything close. So while the interview loop looks like a standard bank loop (an online test, a recorded video, a Superday), what it's actually checking is whether you can write correct software inside a regulated trading business, and for the quant-facing seats, whether the math is a reflex.

## Software engineering and strats are two different interviews

RBC hires technologists into two lanes, and confusing them is the most common way people mis-prep. Capital Markets Technology, what most candidates mean by "tech," builds the platforms: trade booking and order management, pricing and market-data services, risk and regulatory reporting, and the infrastructure the floor runs on. Strats (RBC also lists them as quantitative analysts or quant developers) sit next to the desk. They build the pricing and risk models, then push those models into production and support the traders who use them. A strat is a quant who ships code, and both halves get graded.

The loops share a coding core and split after that. An engineering loop leans on data structures, object-oriented design (this is a heavy Java shop), a system-design conversation for anyone past the new-grad line, and a real behavioral read. A strat loop keeps the coding and layers on probability, expected value, and a math discussion tied to pricing or statistics. Rusty on brainteasers but strong at building things? Aim at engineering. Like the math and want a seat near the trading book? Strats, and the desk-facing quant seats generally pay more because their comp tracks the business they support.

## The process: online assessment, a HireVue, then a Superday

For most technology roles the first gate is an online coding assessment, typically on HackerRank or CodeSignal, pushed to you once a recruiter clears your application. Expect two or three problems in roughly 60 to 90 minutes, auto-graded against hidden test cases. Quant and strat postings often bolt probability or math items onto the same test, so a coding-only study plan quietly leaves points on the table.

Early-career candidates almost always hit a HireVue after that: a recorded, one-way video with a handful of behavioral prompts you answer on your own schedule, sometimes with a light technical question dropped in. RBC uses these heavily across its student and new-grad funnels, including the Amplify summer internship in Toronto. Treat it as a real round even though nobody is on the other end, because a thin "why RBC, why capital markets" answer ends the process here without anyone telling you why.

Clear those and you move to one or two live technical screens with an engineer or quant, 45 to 60 minutes each in a shared editor, split between a coding problem and a walk through your resume. The final stage is a Superday: virtual or onsite in Toronto or New York, three to five back-to-back sessions covering coding, system design, behavioral, and a probability block for strats. Campus pipelines sometimes run this last stage as an assessment centre with group and case elements mixed in.

| Interview stage | Format | What it evaluates | Typical length |
| --- | --- | --- | --- |
| Online assessment | HackerRank or CodeSignal, 2 to 3 coding problems; quant/strat roles add probability and math items | Data structures, correctness against hidden tests under a timer | 60-90 min, auto-graded |
| HireVue video (early-career) | Recorded one-way video, behavioral prompts, occasional light technical question | Communication and motivation ("why RBC, why capital markets") | 20-30 min, self-scheduled |
| Technical phone screen | 1 to 2 live sessions with an engineer or quant in a shared editor | Coding, Java/OOP reasoning, resume depth | 45-60 min each |
| Superday / assessment centre | 3 to 5 back-to-back interviews, virtual or onsite in Toronto or NYC | Coding, system design, behavioral; probability for strats | Roughly half a day |

## What the coding rounds actually ask

The bar is not FAANG-brutal, and pretending otherwise wastes your prep. Think LeetCode easy to medium, tilted toward arrays, strings, and hash maps, with the occasional two-pointer, sliding-window, or light dynamic-programming problem. Because the online round is auto-graded, an off-by-one or an unhandled empty input costs you the same points a human screener might have waved through, so get a correct brute force compiling first and optimize only if the clock allows. A working O(n log n) beats a clever O(n) that dies on one hidden case.

The live rounds drift somewhere a pure algorithm grinder skips, because the stack is Java. Interviewers ask how a HashMap handles collisions, when you'd reach for an interface over an abstract class, or how you'd model a handful of classes for an order book or a trade. The problem phrasings you'll actually hear:

- "Given daily prices in an array, find the max profit from one buy and one sell."

- "Check whether a string of brackets is balanced."

- "Count the pairs in an array that sum to a target."

- "Given a stream of trades, keep a running position and flag when it crosses zero."

That last one, a markets wrapper on a textbook problem, shows up more in strat and desk-facing loops. The move underneath is still a hash map or a prefix sum, and nobody expects you to price a derivative to answer it.

## Probability and brainteasers, the strat filter

This block is where candidates who prepped for a quant seat pull away from candidates who ground LeetCode and hoped. Strat interviews reliably include expected-value and probability questions asked conversationally, where how you set up the problem counts as much as the number you land on. A representative set:

- "Expected number of fair-coin flips to see two heads in a row." (It's 6, and they want to watch you build the recurrence.)

- "Roll a fair die until a 6 shows. Expected number of rolls?"

- "Draw two cards from a shuffled deck. Probability both are aces?"

- "A price does a symmetric random walk, up or down a dollar each step. Probability it reaches +10 before -5?"

Conditional expectation and simple recurrences need to be automatic, not things you reconstruct under pressure. If you stall setting up states, that's the single skill to drill before the Superday, because the interviewer is reading your process, not just checking the final digit.

## System design with a markets accent

RBC's design questions steer away from "design Twitter" toward systems that have to be correct and auditable. Design a service that consumes a market-data feed and keeps live positions. Design an order management system and reason through what happens when a message arrives twice or never arrives at all. The vocabulary they're listening for is idempotency, ordering guarantees, and reconciling state after a process crashes, because a dropped or double-counted fill is real money and a compliance call, not a stale feed.

Senior engineering loops still cover ordinary distributed-systems ground (caching, message queues, database selection, consistency trade-offs), but the tell that you've worked in this domain is leading with correctness and recovery before you talk throughput. A candidate who opens with "how do we guarantee we never lose or replay a fill" is speaking the language of the floor, and the interviewer hears it right away.

## What the engineering culture is genuinely like

The Borealis AI connection is what separates RBC from a generic bank-tech job, and recruiters undersell it. Borealis is RBC's applied research institute, with teams across a few Canadian cities working on reinforcement learning, time-series modeling, and NLP. Aiden, that reinforcement-learning execution platform, came out of Borealis working with Capital Markets, and it's a production system carrying live orders, not a pitch deck. As a platform engineer you probably won't do the research yourself, but you may build the pipelines and services that put those models in front of a trader, which is a more interesting problem than most bank plumbing.

The rest of the stack reads like a large bank several years into modernizing: a lot of Java, growing Python anywhere near data or ML, some C++ and kdb+/q in the low-latency and time-series corners, and an ongoing push onto public cloud. There's legacy, and some of your work will be wrapping or replacing systems older than you are. The real constraint is that this is a bank. Change control, audit trails, and regulatory review are walls you can't code around, and releases are more deliberate than at a product company. The trade in return is a reputation, on the technology side at least, for saner hours and a more collegial floor than the US bulge-bracket firms RBC competes with. People who do well here like the depth and the stakes, and don't miss "move fast and break things" when the thing that would break is a trading book.

## Comp: strong, with the bonus doing the real work

Be straight about the money. Bank technology comp trails Big Tech, and the gap runs wider in Toronto than New York because Canadian tech pay sits below US pay to begin with. A Toronto technology analyst's new-grad base tends to land in the mid-to-high five figures in Canadian dollars, associate titles higher, and the equivalent New York seat pays in US dollars at meaningfully bigger numbers for the same level. Treat those as directional ranges, not quotes; they move year to year and by desk, so pull current figures rather than trusting one screenshot.

The bonus is the variable that actually decides your total. It's discretionary, it swings with the firm's year and your desk's year, and it makes up a bigger slice of total comp on the markets-facing and quant side than on pure platform teams. A chunk of a senior bonus can land as deferred RBC stock that vests over several years, which is the structural difference from a Big Tech package where a large share is liquid RSUs from your first vest. So a bank offer and a tech offer can print similar totals on paper while the cash-versus-deferred split, and the volatility, look nothing alike.

Don't anchor on a single data point. Pull current ranges from [levels.fyi](https://www.levels.fyi), cross-check Wall Street Oasis for the banking read on base and bonus by title, and in the offer call ask directly how the bonus divides between cash and deferred stock and what the vesting schedule is. That answer tells you more about the real value of the package than the headline number does, and it's the question most candidates forget to ask.

## How to prep without burning weeks

Do your problems timed, in the actual assessment tool (HackerRank or CodeSignal), not untimed in a comfortable editor. The interface, the grader, and the hidden test cases are the thing being measured. Cover arrays, strings, hash maps, two pointers, sliding window, and basic dynamic programming, and practice landing a correct answer fast rather than chasing the optimal one against a timer. Since the live rounds lean Java, refresh object-oriented design and the standard collections so you're solving the problem instead of fighting syntax.

Going for strats or quant dev? Spend at least as much time on probability as on code. Expected value, conditional probability, and simple recurrences should be muscle memory. For system design, read how order management systems handle message ordering and failure so you can talk about idempotency and reconciliation without hand-waving. And write a genuine answer for why RBC and why capital markets, because behavioral rounds carry more weight at a bank than at a startup. Being able to say one accurate sentence each about Borealis AI and Aiden is a cheap, effective signal that you did the homework the rest of the field skipped.