You can build a flawless distributed system on the whiteboard and still fail Amazon's system design round. That's because you're being scored on two things at once: the architecture, and how you behave while you draw it.
Amazon weaves Leadership Principles into the technical round, runs a trained Bar Raiser through your loop, and calibrates the same question differently depending on whether you're interviewing at L4, L5, or L6.
The result is a round that rewards engineers who design for failure, talk through their trade-offs out loud, and sound like an owner the whole way through.
This guide covers the system design round specifically: where it sits in the loop, what interviewers actually score, the questions they ask, and how to prepare for the parts that catch people off guard.
Amazon's system design rounds
System design shows up once you reach the onsite loop. Before that, expect a recruiter screen and, for some teams, a quick hiring-manager fit conversation. Junior candidates (SDE I and II) sometimes get an online assessment first; senior candidates usually skip it.
The onsite loop typically runs:
- Two coding rounds
- One or two system design rounds (high-level design, sometimes plus a low-level design round)
- A Bar Raiser round, often scheduled last
For junior SDE loops, the "system design" round is frequently a low-level design (LLD) round built around object-oriented design rather than distributed-systems scale.
Senior loops lean toward high-level design (HLD), and L6 candidates often see dedicated behavioral and system design rounds spread across more than one day.
One thing to internalize early: Amazon interviewers pick their own questions, and the process is team-dependent. You usually interview with people who'd actually be on your team, so the problems skew toward what that team really builds.
What to expect
The round is conversational, not a silent whiteboard exercise. Interviewers want to brainstorm with you, push on your choices ("why this database and not that one?"), and watch how you respond when challenged.
One hiring manager described his ideal candidate as someone he can argue architecture with the way he would with a teammate.
Two formats come up often. The first is a classic open-ended prompt where you architect a system from scratch in about 45 to 60 minutes. The second is what one interviewer calls "reverse system design": instead of handing you a generic prompt, they ask you to walk through a system you've actually built, then start poking holes.
What happens if this dependency goes down? Do you have a circuit breaker? What stops one slow service from hogging every thread? It's hard to fake, because it runs on your real experience.
Whichever format you get, the senior interviewers we spoke with want a structured progression: clarify functional and non-functional requirements first, sketch the API, define your data model and entities, then talk trade-offs and bottlenecks as you go.
Candidates who jump straight to implementation details send a junior signal.
What Amazon evaluates
The job description won't print the real rubric. Here's what the interviewers we talked to are actually scoring.
Resiliency and availability. This is the dimension Amazon interviewers raise more than any other. As one hiring manager put it, "resiliency is baked into every system," and it's where candidates without hyperscale experience can still score points. They're watching for redundancy, failure isolation, circuit breakers, and graceful degradation, not just throughput numbers.
Trade-off articulation. "If you can articulate that trade-off to me, you'll get 100% from me," one interviewer said. There's no perfect system, and they know it. They want to hear why you picked Redis over DynamoDB, and whether you can frame a decision as a one-way door (hard to reverse, so deliberate) versus a two-way door (reversible, so move fast).
Structured thinking and communication. Putting the right boxes on the board isn't enough if you go quiet. Interviewers need you to narrate your reasoning so they can steer, dig into a component, and follow your logic. Keep it interactive.
Security-first thinking. Especially in AWS orgs, strong candidates raise access control and least privilege while they're still gathering requirements, not after the design is done.
Scope, as a level signal. The same question is calibrated by level. An L5 is expected to assemble the building blocks correctly. An L6 is expected to go further: think about access control, mutual exclusivity between features, how QA will test the thing, the business reason behind it, and impact that reaches beyond a single team.
Leadership Principles and the system design round
This is the part that surprises engineers coming from other big tech companies. At Amazon there is no standalone "Leadership Principles round." Instead, the principles are scored inside every round, including system design. As one Bar Raiser told us plainly, "LPs are part of every round at Amazon.
There's no LP round." A recent L4 candidate confirmed the same thing from the other side of the table: their system design interview "had a section where the interviewer asks behavioral questions."
So while you're designing, you're also being read for Ownership, Customer Obsession, and the rest. One SDM described the principles as the company's daily working grammar.
Phrases like "have backbone, disagree and commit" and "one-way versus two-way door" aren't interview trivia; they're how decisions get made on the job, and interviewers expect you to reason that way live.
Two structural features make this concrete.
First, the Bar Raiser: a trained interviewer pulled from outside the hiring team who sits in the loop and leads the debrief, with a strong say in your level.
Second, the debrief and leveling itself. The Bar Raiser we spoke with runs consensus meetings with the hiring manager, and a small share of candidates get down-leveled (roughly 10 to 15% in their experience) when the system design or behavioral signal lands below the bar they interviewed for.
Amazon system design questions
Every question below traces to an Amazon interviewer or candidate we spoke with, or to our Amazon system design question bank.
Amazon interviewers choose their own prompts, so treat these as representative, not a fixed list.
Design an A/B experimentation platform. A Bar Raiser's favorite. You split traffic across experiments, with an assignment service holding the core logic and a streaming pipeline gathering metrics. At L6, they expect you to add access control (no one else should touch your experiment's traffic), mutual exclusivity so a user doesn't land in contradictory experiences, and manual overrides so QA can force test users into a variant.
Manage capacity for a network of delivery lockers (high-level design). A real last-mile problem one interviewer likes to use: given a bank of mixed-size locker slots, design how to allocate capacity to minimize cost per package. Strong answers reason about the heuristic probability of a delivery date and the bottlenecks across each hop from fulfillment center to delivery station.
Walk through and harden a system you built (reverse system design). Rather than a generic prompt, the interviewer asks you to explain your own architecture, then probes failure modes: if an upstream dependency breaks, does the whole site go down, or do you have circuit breakers and isolation in place?
Design a chess game (low-level design). A common object-oriented design prompt: single-player with difficulty levels, then multiplayer. The focus lands on your move-traversal algorithm, your data structures, and how clients stay in sync (websockets versus server-sent events).
Design a discount and coupon system (low-level design). Reported by a recent L4 candidate. You model a class hierarchy for percentage-based and fixed-value discounts and handle the messy edge cases of stacking multiple coupons across multiple items.
For more company-specific prompts (including Design Amazon Prime Video, Design a rate limiter, Design a distributed cache, and detecting fake reviews) work through our Amazon system design question bank.
How to prepare
- Build a story bank for the Leadership Principles and rehearse it as seriously as your architecture, because you'll field behavioral questions inside the design round itself. Our Amazon software engineer interview guide maps the principles to the loop.
- Design for failure before you design for scale. Lead with redundancy, circuit breakers and reliability strategies so resiliency is the spine of your answer, not a footnote.
- Prepare to defend your own systems. Pick two or three things you've built and know, cold, why you chose each component and what breaks first under load.
- Open every problem by clarifying functional and non-functional requirements, and name security explicitly while you're at it.
- Practice both high-level and low-level design. If your loop includes an LLD round, drill object-oriented design and class modeling, not just distributed systems.
- Narrate trade-offs continuously, and reach for one-way versus two-way door language to show you know which decisions deserve more caution.
- Run timed mock interviews so the conversational, get-interrupted rhythm feels normal. Our system design mock interviews put you in front of that pressure before the real thing.
Common mistakes
Trying to "win" the disagreement. Interviewers told us candidates (often from companies with a more combative culture) over-optimize for proving they convinced everyone. Nobody is scoring who won. They want to see how you weighed the options and committed.
Waiting for the interviewer to set scope. The most impactful fix one Bar Raiser named was asking more clarifying questions upfront. Silence here reads as a lack of ownership.
Jumping to implementation too early. Writing out CRUD endpoints or naming your language before the high-level design exists is a junior tell. Establish the shape first.
Designing for scale and forgetting resiliency. You may never have worked at Amazon scale, and that's fine. Skipping failure handling is not. Resiliency is where you score the points scale can't.
Treating behavioral as separate. Vague "we did this" stories, or examples where you surprised your manager at the last minute, undercut the Ownership signal the interviewer is reading even during the design discussion.
FAQs
How long is Amazon's system design interview?
Plan for about 45 to 60 minutes per round. Senior loops, especially at L6, often include dedicated system design and behavioral rounds spread across more than one day.
What system design questions does Amazon ask?
A mix of high-level design (an A/B experimentation platform, locker capacity, Design Amazon Prime Video, a rate limiter) and, particularly at junior levels, low-level object-oriented design (a chess game, a coupon system). Because interviewers choose their own questions, many are mapped to the hiring team's real work.
How is Amazon's system design interview different from Google or Meta?
Three ways stand out. Problems are often practical and team-specific rather than generic, including a "reverse system design" format where you defend your own architecture. Resiliency is weighted more heavily than raw scale. And Leadership Principles are scored inside the technical round, with a Bar Raiser in the loop, rather than in a separate behavioral block.
Does the system design round vary by level?
Yes. Junior loops often replace distributed-systems design with a low-level, object-oriented design round. At L5 you're expected to assemble the right building blocks; at L6 you're expected to add security, testing, business context, and org-level impact to the same problem.
Do I need to design at Amazon scale?
No. Interviewers told us they empathize that most candidates haven't seen that scale, so they dig into resiliency instead, which is a core construct of nearly any system you've worked on.
Will I get a high-level or low-level design round?
Either, and sometimes both. Some loops, especially at junior levels, include a low-level design round focused on OOP and class-based design alongside or instead of high-level system design. Ask your recruiter.
Can I use a SQL database in my design?
Yes. Database choice isn't a scoring factor on its own, even though much of Amazon runs on NoSQL. What matters is that you can justify the trade-off for the problem in front of you.
Resources
- System design interview course for the end-to-end framework
- Amazon system design question bank for company-specific practice
- System design mock interviews to rehearse the conversational round under time pressure
- Amazon software engineer interview guide for the full loop and Leadership Principles prep
Everything you need to know for your next interview
Create your free Exponent account and learn how to ace your interviews.
Get started free →Learn everything you need to ace your system design interviews.
Exponent is the fastest-growing tech interview prep platform. Get free interview guides, insider tips, and courses.
Create your free accountRelated Blog Posts
Google System Design Interview (2026 Guide)

Apple System Design Interview (2026 Guide)

Netflix System Design Interview (2026 Guide)


