Amazon Senior Software Engineer Interview Questions

1.Tell me about a time a customer-facing bug or outage slipped through your team's process and actually reached customers. What did you do, and what changed afterward?

2.Describe a situation where you spotted a serious problem outside your team's charter and decided to own it anyway. What made you step in, and how did it play out?

3.Tell me about a time when you had to deliver a project with significant technical debt. How did you balance shipping vs quality?

4.Describe a situation where you disagreed with your manager or a senior leader. How did you handle it?

5.Tell me about a time you discovered that a deeply-held technical belief you had was wrong. What changed your mind, and how did it affect what you built or recommended afterward?

6.Your team owns a high-traffic Lambda function that processes order events from an SQS queue. Suddenly, the Dead Letter Queue is filling up and order confirmations are delayed for Prime customers. Walk me through how you'd triage this in the first 30 minutes.

7.You're designing the data model for a feature that stores per-user cart state in DynamoDB. The feature needs sub-10ms reads, handles 500K concurrent active carts, and occasionally needs to run a report showing the top 100 most-added SKUs across all carts in near-real-time. How do you approach this?

8.You're running a critical Amazon.com service on ECS with a blue/green deployment setup. You've just deployed a new version, and two minutes after shifting 100% of traffic, you see a 3% increase in 500 errors — not enough to trigger your auto-rollback alarm, but enough to worry you. What do you do?

9.How would you design a recommendation system for an e-commerce platform?

10.Explain how you would implement a distributed locking mechanism for a microservices architecture.

11.Design the session management service for Amazon.com — the thing that keeps you logged in across devices. It needs to handle sign-ins across hundreds of millions of customers and survive an AZ failure without logging anyone out.

12.Walk me through how you'd design the delivery ETA system for Amazon.com — the 'Get it by Wednesday' promise shown on every product detail page. It's read by hundreds of millions of product page loads a day.

13.Design a real-time inventory reservation system for Amazon.com that guarantees that if two customers simultaneously add the last unit of a product to their cart and proceed to checkout, exactly one of them completes the purchase. How do you make this work at Amazon's checkout scale without making every add-to-cart feel slow?

14.Design the search autocomplete service for Amazon.com — the dropdown that appears as you type in the search bar. It needs to return results in under 50ms for every keystroke from customers globally, and the suggestions need to reflect trending searches within roughly 15 minutes of them trending.

15.Design a distributed system for processing millions of orders per day with high availability.

16.Amazon needs to build a unified event bus that every team's service can publish to and subscribe from — something like an internal EventBridge but with stronger delivery guarantees, a global schema registry, and the ability to replay any event from the last 7 days. Design it. How do you prevent it from becoming a single point of failure for every team in the company?

17.Amazon Go stores use computer vision to charge customers automatically when they pick up items. Design the backend billing reconciliation system that takes raw 'item picked up / put back' events from hundreds of stores and produces a final, accurate customer charge — knowing that the CV system has a roughly 2% error rate and customers can dispute charges.

18.Tell me about a time you identified a gap in your team's operational health — something like alarming, runbooks, or on-call load — and fixed it without being asked. What was the before and after?

19.Tell me about a time you had to make a call that would slow your team down in the short term — a refactor, a migration, a hard dependency you refused to take — because you believed it was the right long-term decision. How did you sell it?

20.Tell me about a time you mentored a junior engineer. How did you help them grow?

21.Describe a time you owned a feature or project end-to-end but realized partway through that the original requirements were going to produce the wrong customer outcome. What did you do, and how did you navigate changing course mid-execution?

22.Tell me about a time you were the de facto technical lead for an initiative that crossed multiple teams, but you had no headcount authority over anyone. How did you get alignment on technical decisions when teams had conflicting priorities?

23.Tell me about a time you concluded that a widely-adopted internal tool, platform, or process your organization relied on was fundamentally the wrong solution — not just suboptimal, but wrong. How did you make the case, and what happened?

24.Your team's service is directly in the critical path of a major new Amazon initiative — say, a new Prime benefit or a new seller program — and you're being asked to commit to a launch date that you believe will require your team to take on unsustainable technical risk. Leadership is excited and there's a lot of momentum. What do you do?

25.Amazon Prime's same-day delivery 'order cutoff time' — the deadline by which you have to order to get delivery today — varies by zip code, warehouse proximity, and time of day. Estimate how many distinct cutoff time values Amazon needs to maintain across the US.

26.Estimate the annual AWS compute cost Amazon.com itself pays — the retail side, not AWS's external customers — to serve the product detail page. Walk me through how you'd think about it.

27.Amazon's retail catalog has hundreds of millions of ASINs. A new business requirement says that any ASIN with active customer orders in the last 180 days must have its product compliance attributes re-validated within 72 hours of any regulatory change in its product category — and there are thousands of product categories. How do you design a system that meets this SLA reliably, and how do you know when you're not meeting it?