Amazon Software Engineer Interview Questions

1.Tell me about a time you discovered a bug or product issue that was directly hurting customers, even though fixing it wasn't technically your job. What did you do?

2.Tell me about a time a customer — internal or external — came to you with a frustrating experience that wasn't getting fixed. How did you figure out the real problem, and what did you actually do about it?

3.Tell me about a time you realized the technical approach or mental model you'd been using was actually wrong or outdated. How did you find that out, and how did it change how you work?

4.Describe a situation where you took responsibility for something that broke in production — even though your team wasn't the direct owner. How did you handle it and what was the outcome?

5.Tell me about a time you noticed something going wrong outside your team's service boundary — a bug, a reliability issue, a bad customer experience — and you decided to do something about it. Walk me through what you did and why.

6.Walk me through a time you replaced a complicated system or process with something significantly simpler. What was the old solution, what did you replace it with, and what resistance did you face?

7.Tell me about a time you inherited or worked with a system that had grown unnecessarily complicated — multiple moving parts, confusing logic, hard to change. What did you do to simplify it, and how did you decide what to cut?

8.Tell me about a time you helped a more junior engineer work through a problem they were stuck on. How did you approach it, and how did you balance helping them versus doing it for them?

9.Tell me about a junior engineer you worked with who was struggling — technically or otherwise. How did you help them, and how did you figure out what kind of help to actually give?

10.You've got a Lambda function that processes S3 events, and it's occasionally timing out. How do you figure out what's going wrong?

11.A team upstream from you just changed the schema of the SQS messages your service consumes. Your service broke in prod. Walk me through how you handle it right now and how you'd prevent it from happening again.

12.Your DynamoDB table is hot — a single partition is getting hammered and you're seeing ProvisionedThroughputExceededException on reads. What do you do?

13.You push a CDK deployment to a staging environment and the CloudFormation stack rolls back with a generic 'Resource creation failed' error. How do you find out what actually went wrong?

14.You're building a feature for Amazon.com order history that needs to query across a DynamoDB table partitioned by user_id — but a new requirement asks you to also support queries by order date across all users. How do you handle this without redesigning the table?

15.You're running an A/B test on a checkout flow change and the experiment metrics look positive, but your P99 latency went up 40ms in the treatment arm. How do you decide whether to ship it?

16.You're designing the notifications system for an Amazon Marketplace seller — things like 'your order shipped', 'you got a return request', 'your listing was suppressed'. There are 2 million sellers, event volume spikes massively during Prime Day, and sellers need to receive these within seconds. How do you architect this?

17.You're on-call and you get a CloudWatch alarm: your service's error rate just spiked from 0.1% to 8% and it's been climbing for 10 minutes. You haven't deployed anything in 48 hours. Walk me through your first 15 minutes.

18.Your ECS service is autoscaling fine under normal load, but during Prime Day it fell over at 3x traffic even though you had headroom on CPU and memory. What could explain this, and how do you find out which one it is?

19.Design a distributed rate limiter for AWS API Gateway — one that enforces per-customer request quotas (e.g., 1000 req/min) accurately across a fleet of 500 Gateway nodes worldwide, without a single point of failure.

20.You're building the intake pipeline for Kinesis records from Amazon fulfillment centers — package scan events at ~500K events/second globally — and you need to guarantee that every event is processed exactly once for billing purposes. How do you approach this, and where does 'exactly once' actually break down?

21.Design a real-time inventory availability service for Amazon.com product pages. Millions of customers are looking at the same hot product during a flash sale — how do you make sure the 'In Stock' badge is accurate without hammering your database?

22.We want to build a 'Recently Viewed' feature on Kindle — the last 50 books a user opened, shown in order, updated in real time across their devices. How would you store and serve that?

23.Design a service that powers the 'Estimated Delivery Date' shown on Amazon.com product pages — the '2-day delivery if you order in the next 3 hours' message. What data do you need and how do you compute it fast enough to show on every page load?

24.We want to add a 'Write a Review' reminder email to Amazon.com — 7 days after a customer receives an order, send them an email for each delivered item. How would you design the scheduling and delivery pipeline for this?

25.Design the backend for Alexa's daily reminder system — users set reminders like 'remind me to take my medication at 8am every day', and Alexa has to fire them reliably even if the user's Echo device was offline when the trigger time hit.

26.Design a Prime membership eligibility service — given a user ID, return whether they're a valid Prime member in under 10ms, called on virtually every Amazon.com page load. How do you build this to be both fast and correct?

27.Design the search autocomplete service for Amazon.com — the dropdown that shows suggested queries as a user types in the search bar. It needs to return suggestions in under 50ms for every keystroke, globally. Walk me through your approach.

28.Design the Kindle sync service — when a customer reads on their phone and then picks up their Kindle device, it should open to the exact page they left off. There are hundreds of millions of Kindle users across multiple device types. How do you build this?

29.Design Amazon Go's checkout system — the part that figures out what items each customer took and charges them when they walk out the door. Focus on the backend pipeline from raw sensor events to a finalized receipt. How do you handle the case where the system isn't sure what a customer grabbed?

30.Design a seller metrics dashboard for Amazon Marketplace — third-party sellers need near-real-time visibility into their sales, returns, and revenue for the last 30 days. There are 2 million active sellers globally, and each has potentially millions of orders in the window. How do you serve this without running full table scans on every dashboard load?

31.Design the backend for AWS Cost Explorer — the service that lets AWS customers see their spending broken down by service, account, region, and tag, with the ability to filter and group across up to 12 months of history. Customers range from startups spending $100/month to enterprises spending $10M/month with thousands of accounts. How do you build this?

32.Design the backend system that decides which fulfillment center to route an Amazon.com order to for picking and shipping. An order may contain multiple items stored in different warehouses, Prime customers expect 2-day delivery, and this decision needs to be made in under 100ms at checkout. How do you approach this?

33.Describe a time you identified a recurring operational problem on your team — something that kept burning people's time in on-call or triage — and drove a fix for it. How did you prioritize it against your other work?

34.Tell me about a time you owned a launch or a significant feature end-to-end — from design through production. Where did it go sideways, and what did you do about it?

35.You're midway through a quarter and it's clear you're not going to hit the deliverable you committed to. How have you handled that situation — what did you do, when, and with whom?

36.The 'Add to Cart' button on Amazon.com has a bug where it occasionally double-submits — customers end up with two units in their cart when they only clicked once. How would you triage this?

37.Estimate how many customer reviews are submitted on Amazon.com on a typical day. Walk me through your reasoning.

38.The click-through rate on Amazon product recommendation carousels — 'Customers who bought this also bought' — dropped 12% last week with no code changes deployed. Where do you start?

39.You're asked to reduce the cost of a high-volume Lambda function that's currently running 1 million invocations per day at 512MB memory and averaging 800ms duration. Where do you start, and what's your target?

40.We want to add a 'price drop alert' feature to Amazon.com — a customer saves an item and gets notified when the price drops by a meaningful amount. Estimate the write volume this system needs to handle on a peak day, and then tell me what the hardest engineering problem is.

41.Amazon sellers can run sponsored ads on search results. The team tells you that average ad revenue per search dropped 8% this quarter, but the total number of searches went up 15%. Walk me through how you'd break down what's actually going on.

42.You've been asked to reduce end-to-end latency for the Amazon.com product detail page by 20% without touching the front-end. You can instrument and change backend services. Where do you start, and how do you know what's actually on the critical path?

43.Imagine you're the engineer who owns Amazon's Buy Box — the 'Add to Cart' selection on a listing with multiple sellers. You find out that for approximately 0.01% of purchase attempts, the wrong seller is charged instead of the seller who won the Buy Box at the time of the click. How do you handle this?