AO
AceOffer

Amazon Interview Process & Rounds

Amazon's process starts with an online assessment. Since January 2026 the coding section is often two tasks in about 100 minutes: an algorithm problem in roughly 40 minutes, then about 60 minutes of debugging an existing repository with an AI assistant that will point at files but will not write the fix. Reported repositories include a movie database app and a Jira-like issue tracker. Some candidates still get two algorithm questions instead. The assessment also has a work simulation, where you rate replies to workplace scenarios, and a work-style survey. Phone screens and onsite coding rounds are mostly LeetCode-style problems and Amazon reskins of them: course schedule and dependency ordering, top-k frequent elements, LRU cache, lowest common ancestor and interval problems. Design rounds are often object-oriented design (Amazon Locker, pizza ordering, a shopping cart and checkout) and sometimes distributed design. Every round also asks Leadership Principle questions, and the bar raiser can spend most of an hour on them; many loops now also ask how you use generative AI, and some ask how you check its output. Applied Scientist loops add an ML breadth screen and ML system design.

Part of the full catalog of Amazon interview questions, covering every round.

Key facts

  • •6 distinct round types
  • •57 questions reconstructed from 542 candidate reports
  • •Reports span Oct 2025 – Oct 2026
  • •Refreshed monthly · last updated October 2026

The Amazon loop, from candidate reports

Amazon's process starts with an online assessment. Since January 2026 the coding section is often two tasks in about 100 minutes: an algorithm problem in roughly 40 minutes, then about 60 minutes of debugging an existing repository with an AI assistant that will point at files but will not write the fix. Reported repositories include a movie database app and a Jira-like issue tracker. Some candidates still get two algorithm questions instead. The assessment also has a work simulation, where you rate replies to workplace scenarios, and a work-style survey. Phone screens and onsite coding rounds are mostly LeetCode-style problems and Amazon reskins of them: course schedule and dependency ordering, top-k frequent elements, LRU cache, lowest common ancestor and interval problems. Design rounds are often object-oriented design (Amazon Locker, pizza ordering, a shopping cart and checkout) and sometimes distributed design. Every round also asks Leadership Principle questions, and the bar raiser can spend most of an hour on them; many loops now also ask how you use generative AI, and some ask how you check its output. Applied Scientist loops add an ML breadth screen and ML system design.

What does Amazon ask in each round?

Amazon interviews span 6distinct round types, shown below. Counts reflect distinct questions per round across the loops we’ve indexed.

LeetCode-style problems and Amazon reskins: course schedule and dependency ordering, top-k frequent, LRU cache, lowest common ancestor, meeting rooms, union-find, islands, plus a Linux find implementation and a rate limiter.
20 questions
Most-reported: Course Schedule and Dependency Ordering (Topological Sort) (16 reports)
Leadership Principle questions in every round, with the bar raiser often spending most of an hour on them: tight deadlines, using generative AI, conflict and disagreement, ownership beyond scope, dive deep, critical feedback and failures.
13 questions
Most-reported: Deliver Results: Tight Deadlines and Missed Commitments (44 reports)
Often two coding tasks in about 100 minutes: an algorithm problem (drone routing around a circular hub, placing a delivery center on a grid, the most money in K consecutive bags), then debugging a repository with an AI assistant. Plus a work simulation and a work-style survey.
11 questions
Most-reported: AI-Assisted Debugging OA: Format, Strategy and Reported Repos (29 reports)
Mostly object-oriented design (Amazon Locker, pizza ordering, shopping cart and checkout, restaurant waitlist, pub/sub notifications, library management, a deck of cards), and a distributed job scheduler.
8 questions
Most-reported: Design Amazon Locker (OOD) (5 reports)
Coding screens: out-of-order package IDs, arranging package weights, LeetCode-style problems such as two sum, and for frontend roles a React grid of wrapping boxes, with Leadership Principle questions alongside.
3 questions
Most-reported: Arrange Packages So Adjacent Weights Sum Below 10 (2 reports)
For Applied Scientist roles: an ML breadth screen (bias and variance, overfitting, regularisation, A/B testing) and ML system design for search ranking and recommendation.
2 questions
Most-reported: Applied Scientist ML Breadth: Bias-Variance, Overfitting and Fundamentals (10 reports)
What passing candidates do
  • •Preparing a set of STAR stories tagged by Leadership Principle. Candidates report a dozen or more stories across the loop, and are often asked what they would do differently, even about projects that went well.
  • •Having a concrete answer about generative AI: how you use it, how you validate its output, and a time it was wrong and how you noticed.
  • •In the AI debugging task, reproducing from the README and reading the failing tests before changing code. Some candidates who passed only part of the tests still advanced.
  • •Talking through complexity and dry-running the code, which interviewers ask for on the LeetCode-style rounds.
Where candidates lose points
  • •Stories that are not technical enough. In one L5 loop an interviewer asked the candidate to swap in a more technical story.
  • •Fixing only what the UI shows in the debugging task. In the issue-tracker repository the tests check the status code and message the API returns.
  • •Running out of time in the debugging task. One candidate recalls that time left over from the algorithm question does not carry over, so plan for the full 60 minutes.

FAQ

How many rounds is the Amazon interview?▾
Amazon's loop spans 6 round types: LeetCode-style problems and Amazon reskins: course schedule and dependency ordering, top-k frequent, LRU cache, lowest common ancestor, meeting rooms, union-find, islands, plus a Linux find implementation and a rate limiter., Leadership Principle questions in every round, with the bar raiser often spending most of an hour on them: tight deadlines, using generative AI, conflict and disagreement, ownership beyond scope, dive deep, critical feedback and failures., Often two coding tasks in about 100 minutes: an algorithm problem (drone routing around a circular hub, placing a delivery center on a grid, the most money in K consecutive bags), then debugging a repository with an AI assistant. Plus a work simulation and a work-style survey., Mostly object-oriented design (Amazon Locker, pizza ordering, shopping cart and checkout, restaurant waitlist, pub/sub notifications, library management, a deck of cards), and a distributed job scheduler., Coding screens: out-of-order package IDs, arranging package weights, LeetCode-style problems such as two sum, and for frontend roles a React grid of wrapping boxes, with Leadership Principle questions alongside., For Applied Scientist roles: an ML breadth screen (bias and variance, overfitting, regularisation, A/B testing) and ML system design for search ranking and recommendation..
What is the Amazon interview process?▾
Amazon's process starts with an online assessment. Since January 2026 the coding section is often two tasks in about 100 minutes: an algorithm problem in roughly 40 minutes, then about 60 minutes of debugging an existing repository with an AI assistant that will point at files but will not write the fix. Reported repositories include a movie database app and a Jira-like issue tracker. Some candidates still get two algorithm questions instead. The assessment also has a work simulation, where you rate replies to workplace scenarios, and a work-style survey. Phone screens and onsite coding rounds are mostly LeetCode-style problems and Amazon reskins of them: course schedule and dependency ordering, top-k frequent elements, LRU cache, lowest common ancestor and interval problems. Design rounds are often object-oriented design (Amazon Locker, pizza ordering, a shopping cart and checkout) and sometimes distributed design. Every round also asks Leadership Principle questions, and the bar raiser can spend most of an hour on them; many loops now also ask how you use generative AI, and some ask how you check its output. Applied Scientist loops add an ML breadth screen and ML system design.
How fresh is this Amazon interview data?▾
It's reconstructed from candidate reports spanning Oct 2025 – Oct 2026 and refreshed monthly as new reports come in.

Get the full Amazon catalog

Every question, every candidate-reported follow-up, and what passers actually do. Monthly refresh.

See pricing← All Amazon interview questions
Is this helpful?