Big tech AI

Amazon

AWS AI services, Alexa and a large applied-science org, with the Leadership Principles running through every interview.

Overview

Amazon's machine learning work is spread across several genuinely different organisations rather than concentrated in one lab. AWS builds and operates the AI services other companies build on, along with the training and inference infrastructure underneath them. Alexa and the devices group work on speech, dialogue and on-device constraints. Retail, advertising, supply chain, fulfilment and robotics each run their own applied science teams on forecasting, ranking, pricing and demand. Amazon Science publishes across all of it, so you can usually read the specific team's work before you meet them.

The hiring process is the most openly documented of any employer in this group. Amazon publishes its Leadership Principles, says plainly that interviews are structured around them, and publishes role-specific preparation guides on amazon.jobs — including ones for applied scientist and research scientist that describe the technical phone screening, the interview loop, and the fact that different interviewers are assigned different competencies to evaluate. Read the guide for your exact role before you do anything else.

The consequence is that presentation is graded, not only substance. Strong engineers fail here for structuring their answers badly against a rubric that was published in advance and that they could simply have read.

What Amazon visibly values

Inferred from public material, not from a leaked rubric. Treat these as the themes worth emphasising rather than as a scoring sheet.

Leadership Principles evidenced rather than named. Dive Deep means you can state the metric, the query you ran and the number before and after, not that you say you like details.
Working backwards from the customer — being able to say which customer problem a model serves before describing its architecture.
Ownership across the whole lifecycle. Applied scientists here are expected to work alongside engineers to get a model into production rather than hand over a notebook.
Frugality treated as an engineering constraint: cost per inference, instance choice, and an honest answer about whether a smaller model would have done.
Structured, legible explanation. Interviewers are assigned specific competencies to assess, so a rambling answer costs you even when the underlying work was good.
Operational reasoning at scale — what breaks at Amazon's volume, and what a migration costs when many teams depend on the interface you want to change.

Roles this kind of team hires

Linked titles have a full question bank with model answers. Unlinked ones are titles this employer uses that do not map onto one of our role guides.

How to prepare

Read Amazon's own prep guide for your role

Amazon publishes role-specific interview preparation on amazon.jobs, including applied scientist and research scientist guides that set out the screening and loop structure. Very few employers hand you this, and not reading it is an unforced error.

Build a dozen real stories, not sixteen invented ones

Write out projects you actually did in STAR form and map each to two or three Leadership Principles, rather than manufacturing one story per principle. Interviewers probe for your specific actions and the measured outcome.

Pick the organisation, then read its science

AWS AI services, Alexa, advertising and supply chain are different jobs with different problems. Find recent Amazon Science publications from the group you are meeting and be ready to discuss one in detail.

Get specific about cost

For any system you propose, know roughly what it costs per request and what you would cut first. Frugality is one of the published principles, so treat cost as part of the design rather than an afterthought.

Practise disagreeing well

Have Backbone; Disagree and Commit is a published principle. Prepare an example where you pushed back on a technical decision, what evidence you brought, and what you actually did once the call went against you.

What to watch for

Common mistakes with this kind of employer

Treating the behavioural questions as a formality. They are assessed against published principles by interviewers assigned to them, and a thin set of stories sinks candidates who are technically strong.
Vague ownership. "We built a model that improved conversion" tells an interviewer nothing about what you did; they are listening for the first person singular and a number.
Preparing for "Amazon" rather than for the team. The gap between an AWS inference role and an Alexa speech role is wider than the gap between two companies.
Reciting principle names. Quoting Customer Obsession back at an interviewer with no example behind it reads worse than never mentioning it at all.

Sources

This guide is written from primary public material — Amazon's own engineering writing, research pages, published job descriptions, and public talks. Nothing here is taken from a competitor's interview guide, and no specific claim about the loop is made without a source that supports it.

Last reviewed 31 August 2026. Interview processes change without notice and vary by team, level, and location. Amazon's own careers page is the authoritative description of how it hires; where this guide and that page disagree, that page is right.

Rehearse it before it counts.

Knowing what a company values is not the same as demonstrating it under time pressure. Book a 30-minute 1:1 and run the round you are least confident about.

Book a mock interview