Amazon Interview Questions
& Process for Software Engineers

Your interviewers each get handed a short slice of the 16 Leadership Principles before they walk in, two or three apiece, and those are the principles they write you up against. So the loop is not one behavioral test with four witnesses. It is four or five separate ones, each opening with 15 to 20 minutes of Leadership Principle questions before anyone mentions the technical problem, and each producing its own LP feedback. Your best story lands once, in the room where you told it, against whichever principle that interviewer happened to draw. Around a quarter of candidates who clear the technical bar still get rejected on behavioral grounds. Four to eight weeks end to end, with the resume screen filtering most applicants before a recruiter reads a line. Amazon lays out the same loop on its own how we hire page.

The Online Assessment sits in front of the loop as a gate of its own. Only one part of it is timed: 90 minutes of coding across two questions, easy to medium, in an editor that checks your code is syntactically valid rather than running it. Around it sit a Work Simulation, a day-in-the-life scenario with no coding in it, and a Work Styles questionnaire of roughly 38 statements that maps you against the Leadership Principles. SDE II candidates get an extra system-design block of 15 to 20 minutes. Most people treat the untimed pieces as paperwork, which is a mistake given what the Work Styles answers are measuring.

There is no L3. The ladder opens at L4 (SDE I), runs through L5 (SDE II) and L6 (Senior), and tops out at L7 (Principal). Loop performance sets your level, and unlike your salary the level itself is not negotiable, which catches out people who assumed both were on the table. Down-leveling is common, and traced back it is almost always a system design round that stayed shallow or LP answers that arrived with no metrics in them.

Scope is the thing that moves the number. Describing what the team shipped without saying what you personally decided reads as thin, and every interviewer in the loop is writing that impression down. A level holds when the system-design round goes deep on one component rather than touring all of them, and when the behavioral answers are still standing after five follow-ups. The figures below are levels.fyi medians; an offer moves up or down that ladder with the scope you show.

Median total compensation per Levels.fyi (US, October 2026); self-reported and updated continuously. Amazon vests RSUs 5/15/40/40 and front-loads a signing bonus, so any single year's pay runs below or above these blended figures.

An L5 loop is four to five rounds of about an hour: two coding, one system design, and one slot that is either a Low-Level Design round or another coding problem. You write in Amazon's internal editor, which will not run what you write. Nothing executes at any point, so the only way anyone finds out whether your solution works is you walking the interviewer through an example out loud.

The spoken dry-run ends up carrying weight that a passing test suite would carry somewhere else, and clean code stops being a courtesy. Difficulty skews to LeetCode medium and heavy dynamic programming is played down, which puts the bar less on knowing an exotic algorithm and more on producing something correct, readable, and complexity-aware on the first pass. One round in the loop is the Low-Level Design interview, where you model classes and apply design patterns for a prompt like Amazon Locker or a parking lot. Google has no equivalent, and candidates who prepped for a Google loop tend to meet it cold.

“Strong Leadership Principle answers will pull a borderline technical candidate over the line. Clean code has never once done the reverse.”

One person in your loop does not work for the team hiring you. The Bar Raiser is trained for the role, reads everyone's feedback ahead of time, and runs the debrief once it starts. They are there for Leadership Principles and culture fit, and they hold a unilateral veto. The mandate is that each hire should come out better than roughly half of current employees at that level, which is why the standard keeps drifting upward.

By the time the debrief opens, every interviewer has already filed written feedback and an explicit hire or no-hire vote, so the room argues from documents rather than memory. The veto gets used less often than its reputation suggests, because a good Bar Raiser would rather talk the group into agreement than overrule it. A single behavioral flag still carries enormous weight, and candidates with clean coding rounds have been turned down over an LP concern.

Amazon will usually not tell you why you were rejected or down-leveled, which makes the written record the only version of you that survives the loop. The interviewer who liked you and filed two thin sentences helps you less than one who leaned negative and filled a page with specifics, because the debrief can only argue from what is on the page. Quantified results in your stories give the people defending you something they can read out loud.

Amazon RSUs vest 5%, 15%, 40%, 40% across four years, so a fifth of the grant is all you see in the first two. The Year-1 and smaller Year-2 sign-on bonuses exist to cover that trough and keep your cash whole until the equity ramps, which is why the headline TC on the offer letter looks so level year to year. That flatness is cash doing the work of equity. Real equity income only climbs in years three and four, assuming the stock holds.

Base salary is capped at $350K, and past that ceiling every extra dollar has to arrive as sign-on or RSUs. That is also where almost all of your negotiating room lives. Your recruiter is working against a per-level total-comp band and can reshuffle the components inside it, so asking for more money in the abstract goes nowhere while asking for a bigger equity grant sometimes works. A competing offer is your strongest lever.

The largest number in any Amazon negotiation is not a component at all. Getting leveled up shifts you onto an entirely higher band, and it is worth more than anything you will win arguing over sign-on. The vesting schedule is fixed and will not move for you no matter how well the conversation goes. And the total on the offer assumes the stock appreciates 10 to 15% a year, which is Amazon's own math, so what they quote and what lands in your account can be different numbers.

Across candidate reports the same short list keeps surfacing, dressed in Amazon's own warehouse and logistics vocabulary. A Premium question arrives as a grid of connected delivery zones. A Premium question arrives as spoilage spreading through a fulfillment center. The reskin is thorough enough that people who prepared by problem title spend the first minutes of a round not recognizing something they have solved dozens of times.

The full set of Amazon coding interview questions is worked end to end, with patterns and complexity, in the guide.