In this article
Before you start
This page sits in the interview loops guide, which shows where company loops fit among other interview types.
Why interviewers ask this
Candidates ask about the process because the stakes of a wrong guess are large: preparing only for algorithm questions, or only for behavioral stories, leaves half of a loop uncovered. Amazon's page says the loop tests "technical and non-technical competencies", based on the Leadership Principles, and that different interviewers are assigned to evaluate each area. A candidate who knows that can split preparation to match.
The process also explains why the interviews feel different from a pure research talk or a pure coding screen. The role is applied: a posting for an applied scientist role asks for experience "building machine learning models or developing algorithms for business application", and Amazon's page describes applied science as "highly experimental" and needing strong theoretical analysis. Interviewers look for scientific depth and for evidence that you shipped work, and the sections below separate what Amazon states from what you must infer.
The applied scientist process from first principles
In plain words
- Interview loop
- An interview loop is a series of individual interviews with current employees that forms the final stage of Amazon's process.
- Applied scientist
- An applied scientist at Amazon is a scientist who builds and tests machine learning methods for use in a product or business process.
Think of the process as a driving test with a written exam and a road test. The phone screen checks that you can answer on the spot. The loop puts you on the road with several examiners, each watching a different skill. The analogy breaks in one place: here the examiners compare notes afterwards.
In one line: the process is an application, a technical phone screen with a senior team member, then a loop of four interviews. Those interviews mix science, coding and behavioral questions. Amazon assigns different interviewers to evaluate different competencies, every interviewer typically also asks behavioral questions, and the debrief combines what they saw. Everything here is on Amazon's pages.
What the official page does not say
Amazon's applied scientist page is a high-level overview. It lists four technical headings, "technical topics", "problem solving and coding", "knowledge depth and breadth" and "tech talk", and gives no detail under any of them. It does not say which languages are allowed, how hard the coding is, which machine learning topics are covered, how long a tech talk lasts or whether the loop includes a presentation.
A tech talk, in general use, is a presentation of your own research or project to a technical audience, followed by their questions. Amazon's page does not describe it. One candidate for a senior applied scientist role reported a 45-minute seminar with a 1 to 2 page abstract sent in advance (Blind, August 2020, fetched 2026-10-24): one report, not Amazon's format.
That silence matters. It means every claim you read elsewhere about the exact topics, order or number of coding rounds is either one candidate's report or an inference. For the missing details, ask your recruiter: Amazon says the process "differs from role to role" (Interviewing at Amazon).
Step 1: Submit the application
The page says that to be considered you "must first submit a job application", and that the application and interview process differs from role to role. The qualifications are in the posting. One posting read for this page asks for a doctorate, or a master's degree and four or more years of science, technology or engineering experience, one or more years of programming in Java, C++, Python or a similar language, and experience with standard tools such as random forests, gradient-boosted regression, LASSO (least absolute shrinkage and selection operator) and logistic regression.
The input is your resume and the posting. The output is an application the team can screen. Write each resume line as a result with a number, because Amazon's page says answers "should include metrics or data where applicable", and the same habit helps here. You now have an application in, and step 2 is the first conversation.
Step 2: Pass the technical phone screen
A technical phone screen is a first interview by phone or video that mixes behavioral and technical questions. Amazon's page says there are one or two, each 60 minutes, with a senior leader on the team, and that the questions cover science, technical topics and the Leadership Principles. The number depends on the team and role. Amazon lists the screen's outcome as its own stage: if the screen is successful, a recruiter contacts you to arrange the interview loop.
The input is a 60-minute conversation; the output is an outcome from the team. Prepare two things for it: one project you can explain at three depths, from one sentence to the full method, and two or three stories from your past work. You now have a decision to continue, and step 3 is the loop.
Step 3: Complete the interview loop
The loop is four interviews of 55 minutes each, with members of Amazon's science community, and the page says they assess technical and non-technical competencies based on the Leadership Principles. The technical headings are the four listed above. The behavioral part is substantial: Amazon says "a significant portion of the conversation will focus on how you've demonstrated our Leadership Principles", that each interviewer will typically ask two or three behavioral questions, and that it will not ask brain teasers.
Bar Raisers appear in loop interviews. According to the About Amazon article, they are volunteer employees who serve as objective third-party advisers, lead the debrief and keep it data-driven, so the decision does not rest on one person. The input is four conversations; the output is a set of written assessments. You now have the evidence the debrief uses, and step 4 is the result.
Step 4: Receive the outcome and the offer discussion
The page says the interview outcome arrives within five business days, and the last stage it lists is the offer discussion. Nothing on the page describes what happens in an offer discussion or how the loop's assessments are weighed, so treat anything beyond "outcome, then offer discussion" as unstated. Ask your recruiter what the offer discussion covers, and what happens if the outcome is a no, because the page is silent on both. The input is the debrief, and the output is a decision you can act on. Plan your notice period and other deadlines around that five-day window. For offer conversations, see the salary negotiation guide.
Step-by-step description
- Submit a job application first.
- One or two 60-minute interviews.
- Four interviews of 55 minutes.
- Outcome in 5 business days, then offer talk.
- The technical phone screen is one or two interviews of 60 minutes each.
- The loop is four interviews of 55 minutes each, followed by an outcome within five business days.
Knowing it is half the job. Saying it is the other half.
Interview Clozo is an AI interview copilot for real interview rounds: it hears the question and puts a strong answer on your screen, and you say it in your own words. See the features or how it works.
Start Free Trial →Worked example
The figures below are the ones Amazon's page states, combined by a short script into totals you can plan around.1
| Fact | Value |
|---|---|
| Technical phone screens | 1 to 2, 60 minutes each |
| Loop | 4 interviews, 55 minutes each |
| Loop time | 220 minutes |
| Total interview time | 280 to 340 minutes (4.67 to 5.67 hours) |
| Behavioral questions in the loop | 8 to 12, at 2 to 3 per interviewer |
| Leadership Principles | 16 |
| Outcome after the loop | Within 5 business days |
One screen of 60 minutes plus the loop's 220 gives 280 minutes, and two screens give 340. Four loop interviewers at two to three behavioral questions each gives eight to twelve questions, before counting the screen.1 The page does not give a question count for the screen, so that part is left out.
You cannot know which of the 16 principles each of those 8 to 12 questions will target, and one real story usually shows more than one principle (a launch you led can show Ownership, Dive Deep and Deliver Results). A small bank of stories, each tagged with two or three principles, therefore covers all 16 with fewer stories than principles, and it adapts to follow-ups in a way sixteen scripted answers do not. See Amazon Leadership Principles for how many stories cover all sixteen in an illustrative case.
Text description of the diagram
- The phone screens take 60 to 120 minutes and the loop takes 220 minutes.
- The total is 280 to 340 minutes.
A plan that follows from these numbers is below. It is advice, not a figure from Amazon, and the week split is illustrative.
| Week | Focus | Output |
|---|---|---|
| 1 | Behavioral stories mapped to the Leadership Principles | A story bank with a number in every result |
| 2 | Machine learning fundamentals and depth on your own work | Answers to "why this model" and "what would you change" |
| 3 | Coding and problem solving | Timed practice in your best language |
| 4 | A short talk on your strongest project, then practice rounds | A talk on your strongest project, sized to the length your recruiter gives, that survives questions |
Trade-offs and edge cases
Amazon's page leaves room for variation, and these are the cases where preparing from a single description misleads.
When the team changes the shape
The number of phone screens is "one to two, depending on the team and role", and the process "differs from role to role". A research-leaning team and a product team can weigh coding, depth and the talk differently. Ask the recruiter which of the four technical headings carries the most weight for your team, and what the tech talk should cover.
Text description of the diagram
- Amazon lists technical topics, problem solving and coding, knowledge depth and breadth, and tech talk.
- Behavioral questions are based on the Leadership Principles.
When the posting asks for more than the page
The interview page names competencies, while the posting names tools. The posting read for this page lists random forests, gradient-boosted regression, LASSO and logistic regression as standard techniques, and optimization and forecasting as preferred. Treat those as likely depth areas for that team, which is an inference, and review them against your own projects. The bias and variance guide and the probability and statistics guide cover the fundamentals.
When coding is a surprise
The page lists "problem solving and coding" and links a sample coding question, so plan to write code. It does not give a language or platform, so ask. The machine learning coding guide covers the model-from-scratch style of problem, and the machine learning system design guide covers design questions that often accompany science loops, though this page cannot say Amazon asks them.
How to answer it in an interview
An illustrative answer (invented project and figures):
Tell me about a complex machine learning project you worked on.
I built a fraud-ranking model for an order-review team. The problem was that manual review could cover only a fraction of flagged orders, so I needed to rank them by risk. I chose gradient-boosted trees because the features were tabular and the team needed explanations, tested against a simple baseline, and measured precision at the review budget. The result was 30% fewer reviews spent on legitimate orders at the same review budget. The part I would change is the labels: they lagged by weeks, so I would add a delayed-feedback check.
The spoken structure is: problem and constraint, method and why, how you measured it, the result with a number, then what you would change. For the stories behind it, see the STAR method guide.
Common mistakes
- "The loop is just coding" (wrong). Amazon says the loop covers technical and non-technical competencies based on the Leadership Principles (right: split preparation between science, coding and behavioral stories).
- "Only the on-site is behavioral" (wrong). The phone screen lists Leadership Principles among its question categories (right: have stories ready from the first call).
- "Memorize an answer per principle" (wrong). Each interviewer asks two or three behavioral questions, and a rehearsed script breaks on follow-ups (right: build a small story bank and adapt it).
- "Describe what the team did" (wrong). The page says answers should show your experience and decisions (right: say what you did, what you decided and why, with a metric).
- "Assume the tech talk is a quick demo" (wrong). The page gives no format (right: ask your recruiter, and prepare a project you can explain at several depths).
Follow-up questions
How to crack the Amazon applied scientist interview?
Cover the four areas Amazon lists, in proportion: technical topics, problem solving and coding, depth and breadth of knowledge, and the tech talk, plus behavioral stories for the Leadership Principles. Ask your recruiter what the tech talk should contain. No page promises an outcome; preparation raises your odds, not certainty.
How to prepare for the Amazon applied scientist interview?
Build a story bank first, then review the machine learning methods named in your target posting, practise coding in your best language under time, and prepare one project to explain at three depths. The interview loops guide shows how this fits with other company loops.
Explain a complex ML project you have worked on?
Name the problem and its constraint, the method and why you chose it over a baseline, how you measured it, the result with a number, and what you would change. Keep each part to a few sentences so follow-up questions can take the depth.
Does Amazon ask coding questions in the applied scientist loop?
Amazon's page lists "problem solving and coding" as a technical competency and links a sample coding question, so yes, prepare to code. It gives no language or difficulty, so confirm both with your recruiter.
What does a Bar Raiser do in an Amazon interview?
According to Amazon, a Bar Raiser is a volunteer employee who acts as an objective third-party adviser, leads the debrief and keeps it evidence-based, so that no single person makes the hiring decision. The Leadership Principles guide covers how interviewers use the principles.
- Amazon's page lists an application, one or two 60-minute technical phone screens, a loop of four 55-minute interviews, and an outcome within five business days.
- The loop mixes technical headings with behavioral questions based on the 16 Leadership Principles.
- The page gives no detail on topics, coding difficulty or the tech talk; ask your recruiter.
- Prepare a story bank with a number in every result, and one project you can explain at three depths.
- Treat third-party reports as one candidate's experience, and check them against the official page.
Sources
- Applied scientist interview prep — Amazon Jobs (no page date shown; date = read date), 24 Oct 2026
- Interviewing at Amazon — Amazon Jobs (no page date shown; date = read date), 24 Oct 2026
- In-person interview: what to expect — Amazon Jobs (no page date shown; date = read date), 24 Oct 2026
- Amazon Bar Raiser — About Amazon (no page date shown; date = read date), 24 Oct 2026
- Amazon leadership principles — Amazon Jobs (no page date shown; date = read date), 24 Oct 2026
- Amazon Sr Applied Scientist presentation (candidate report) — Blind (post dated August 2020; date = fetched date), 24 Oct 2026
- Applied Scientist, Machine Learning, Amazon Transportation (job 10496052) — Amazon Jobs (posting shows no date; date = read date), 24 Oct 2026
