Interview prep · last verified 2026-06-08

Amazon interview prep: the real loop

A public-sources breakdown of Amazon's software interview loop — the rounds you face, the competencies each one scores, and the question themes to expect. Then practice a live AI-avatar mock calibrated to Amazon, and get a hiring-committee hire/no-hire verdict.

Run a live recruiter screen calibrated to Amazon — your first is free.Start free →

Software Engineer roles

Amazon L5 — the loop

  1. Recruiter Screen30 min · Technical recruiter

    Band mapping: this KB's L5 corresponds to Amazon SDE III / Senior SDE (Amazon internal L6). Recruiter sets expectations on Leadership Principles coverage and STAR-format answers; LP stories are the backbone of every onsite round. Per Amazon's official SDE III prep page, a 60-minute technical phone screen (30 min Leadership Principles + 30 min coding/system design) precedes the loop, and the official loop is five 55-minute interviews that include coding round(s) — coding is not simulated in this product.

  2. Behavioral Deep-Dive55 min · Senior engineer trained as an interviewer, often outside the team

    One to three assigned Leadership Principles per interviewer (commonly two or three), probed via STAR stories with heavy data follow-ups; Amazon's official guidance says each interviewer typically asks two or three behavioral questions. The Bar Raiser coordinates LP coverage across the loop so principles are addressed without excessive overlap.

  3. System Design55 min · Senior or Principal engineer

    Design round still carries assigned LPs (commonly themes like ownership and operational excellence woven into design probing). Expect operational readiness questions: metrics, alarms, deployment safety. Official SDE III guidance: expect at least one system design question, judged on practicality, scalability, and reliability.

  4. Hiring Manager Round55 min · Hiring manager

    Mix of role fit and assigned LPs. In the debrief, Amazon officially describes the Bar Raiser and hiring manager as jointly driving the final decision; public guides describe both as the only interviewers with effective veto power.

  5. Leadership / Cross-functional55 min · Bar Raiser: trained interviewer from an unrelated team

    Bar Raiser interview (modeled here as a distinct round; officially the Bar Raiser is one of the loop interviewers). Objective per Amazon: every hire should be better than 50% of current Amazonians in similar roles. This interviewer is independent of the hiring team, digs hardest on data and contradictions across stories, and drives the final decision jointly with the hiring manager — widely described as an effective veto.

What each round scores

  • Ownership: Acting on behalf of the whole company, beyond own team scope; never saying 'that's not my job'; owning outcomes long-term.
  • Dive Deep: Operating at all levels, staying connected to the details, auditing frequently, and being skeptical when metrics and anecdotes differ.
  • Deliver Results: Focusing on key inputs and delivering with the right quality and in a timely fashion despite setbacks.
  • Insist on the Highest Standards / operational excellence: Relentlessly high standards: defect prevention, operational rigor, and refusing to ship known problems downstream.
  • Bias for Action and frugality: Valuing calculated speed: many decisions are reversible and don't need extensive study; accomplishing more with less.

Question themes to expect

  • Ownership beyond assigned scope: Tell me about a time you took on a problem that was clearly outside your job's boundaries. What happened?
  • Having backbone, then committing fully: Describe a decision you strongly disagreed with that your team made anyway. What did you do before and after the decision?
  • Diving deep when the data looks wrong: Tell me about a time a metric or report didn't smell right to you, and you dug in personally. What did you find?
  • Delivering under pressure without lowering the bar: Tell me about your tightest deadline on something that mattered. How did you deliver, and what did you refuse to compromise?
  • Working backwards from a customer problem: Give me an example of a technical decision you changed because of something you learned about customers.

Amazon L6 — the loop

  1. Recruiter Screen30 min · Technical recruiter

    Band mapping: this KB's L6 corresponds to Amazon Principal Engineer (Amazon internal L7). Recruiter probes for principal-shaped evidence: influence across an org, technical strategy artifacts, operational leadership at scale. Note: Amazon publishes no Principal-specific prep page; this loop extrapolates the verified senior pattern (4-6 interviews, LP-driven, Bar Raiser included) with public reports of heavier behavioral/strategic weighting at L7.

  2. Behavioral Deep-Dive60 min · Principal engineer from another org

    LP coverage at principal calibration: Think Big, Are Right A Lot, Hire and Develop the Best feature heavily. Stories must show multi-team scope and multi-year horizon.

  3. System Design60 min · Principal or Senior Principal engineer

    Architecture strategy: large-scale systems with explicit cost models, operational ownership across many teams, and evolution under growth. Expect deep frugality and operational-excellence probing.

  4. Hiring Manager Round60 min · Hiring manager (typically a director or senior manager)

    Charter fit and LP coverage. At principal level the conversation centers on how the candidate would move the org's hardest technical problems.

  5. Leadership / Cross-functional60 min · Bar Raiser, senior and outside the org

    Bar Raiser at principal calibration: cross-examines for consistency across the loop, probes whether scope claims survive detail-level questioning (Dive Deep applies to the candidate's own claims). Final decision is driven jointly with the hiring manager. One ex-Bar-Raiser account of debrief mechanics: the Bar Raiser polls the most junior interviewer first so senior votes don't anchor the room, and opens by asking each interviewer whether any answer was genuinely impressive — absence of excitement is treated as data (one practitioner's technique, not stated policy).

What each round scores

  • Think Big: Creating and communicating a bold multi-year direction that inspires results: looking around corners for the org.
  • Are Right, A Lot: Strong judgment with good instincts: seeking diverse perspectives and working to disconfirm own beliefs on consequential calls.
  • Hire and Develop the Best: Raising the performance bar with every hire and promotion; developing leaders and taking the role of multiplier seriously.
  • Architecture and operational strategy at org scale: Designing systems and operational models that dozens of teams can build on and run safely, with explicit cost structures.
  • Earn Trust across organizations: Building credibility with teams and leaders who don't report to you, through candor and follow-through.

Question themes to expect

  • Setting bold multi-year technical direction: Tell me about the biggest technical bet you have convinced an organization to make. Take me from idea to funded plan.
  • Judgment quality on consequential calls: Pick the highest-stakes technical decision you have made. Convince me your process was sound, independent of the outcome.
  • Raising operational standards across many teams: Tell me about a time you raised the operational bar for an entire organization, not just a team. What did you change and what did it cost?
  • Architecting with cost as a first-class constraint: Design a multi-tenant event processing platform for an org of 30 teams where infrastructure cost is under executive scrutiny.
  • Developing the senior engineering bench: Tell me about an engineer you developed into senior or principal scope. What exactly did you do that mattered?

Amazon L7 — the loop

  1. Recruiter Screen45 min · Senior technical recruiter or executive recruiter

    Band mapping: this KB's L7 corresponds to Amazon Senior Principal Engineer (Amazon internal L8). Screens are evidence-heavy: company-scale technical influence, published artifacts, multi-org programs. Note: no official prep material exists for this level; loop shape extrapolated from the verified senior pattern and principal-level public guidance. Practitioner accounts describe leveling along four dimensions — scope, contribution, impact, difficulty. The Principal bar: work that affects organization-wide operations, 'we'-led initiatives requiring multi-team coordination, strategic (not task) impact, and managing fundamental trade-offs between competing organizational needs.

  2. Behavioral Deep-Dive60 min · Senior Principal engineer or Distinguished Engineer

    LP coverage at senior-principal calibration: stories must show influence across business lines, multi-year horizons, and decisions affecting hundreds of engineers.

  3. System Design60 min · Senior Principal or Distinguished Engineer

    Company-scale architecture strategy: portfolio of systems, build/buy/standardize calls, and the operating model for technology shared across business units.

  4. Leadership / Cross-functional60 min · Bar Raiser, senior, from a distant org

    Bar Raiser round: hardest scrutiny on whether claimed scope was truly the candidate's, with detail-level probing to separate proximity to big things from authorship of them.

  5. Hiring Manager Round60 min · Director or VP

    Charter and mutual evaluation: which company-level problems the candidate would own, and their thesis for them. Expect Working Backwards thinking applied to their own role.

What each round scores

  • Company-scale technical vision: Setting direction that crosses business-unit boundaries: multi-year bets affecting how the company builds, with executive sponsorship.
  • Are Right, A Lot at irreversible scale: Judgment on one-way doors where the company cannot easily recover: platform standardization, deprecation of business-critical systems, major build/buy.
  • Mechanisms over good intentions: Building self-reinforcing processes (reviews, metrics, forcing functions) that make an org reliably do the right thing without the candidate present.
  • Hire and Develop at organizational scale: Shaping the senior engineering population: principal bench building, bar ownership across an org, succession for their own role.
  • Customer obsession at strategy level: Anchoring company-scale technical strategy in customer outcomes rather than technology fashion.

Question themes to expect

  • Authoring a company-scale technical bet: Tell me about the largest technical direction you have authored: one that crossed organizational boundaries and took years. I'll want the paper trail.
  • Irreversible decision at company scale: Walk me through the most irreversible technical decision you have owned. How did you decide, and how do you score it now?
  • Building mechanisms that outlive attention: Good intentions don't work; mechanisms do. Tell me about the best mechanism you have built, and prove it still works without you.
  • Standardizing technology across resistant business units: Two business units run rival internal platforms doing the same job, each with executive backing. You're asked to drive a resolution. What do you do?
  • Deprecating a system tied to live revenue: Tell me about a time you championed retiring a system that was still making money. How did you make the case and land the migration?

Machine Learning Engineer roles

Amazon L5 — the loop

  1. Recruiter Screen30 min · Technical recruiter (ML roles)

    Band mapping: this KB's L5 corresponds to Amazon Senior Applied Scientist / Senior MLE (internal L6) depending on the team's ladder. Recruiter clarifies applied-scientist versus SDE-ML track, since loops differ in science depth. Per Amazon's official Applied Scientist prep page, the loop is four 55-minute interviews with the science community, covering ML depth/breadth, problem solving and coding, a tech talk, and LP-based behavioral questions; this simulation maps those onto the rounds below and does not cover coding or the tech talk. Officially, the loop is preceded by one or two 60-minute technical phone screens conducted by a senior leader (science depth, technical questions, and Leadership Principles in the same call), and the outcome is communicated within roughly five business days.

  2. Behavioral Deep-Dive55 min · Senior applied scientist or MLE, LP-assigned

    LP-driven STAR rounds identical in format to SWE — official guidance: each interviewer typically asks two or three behavioral questions; ML stories must still produce numbers under Dive Deep: dataset sizes, metric deltas, revenue or cost impact. Practitioner hiring-manager guides add data-literacy and evaluation probes to this round: a misleading summary statistic the candidate caught, unexpected or biased model outputs and the guardrails around them, and what they do when performance breaches a threshold.

  3. System Design55 min · Senior MLE or applied scientist

    ML system design with operational and cost emphasis: training pipelines, real-time inference economics, model monitoring. Frugality is probed explicitly — inference cost per request is fair game. In the official loop this maps to the ML depth/breadth and design competencies.

  4. Hiring Manager Round55 min · Hiring manager

    Role fit plus assigned LPs; expect discussion of how the candidate ties model work to business metrics, a strong Amazon expectation.

  5. Leadership / Cross-functional55 min · Bar Raiser from an unrelated team

    Bar Raiser: may not be an ML specialist, which itself is a test — candidate must explain ML decisions in business and customer terms while surviving generic Dive Deep probing. Final decision is driven jointly with the hiring manager.

What each round scores

  • Customer-obsessed ML framing: Starting from the customer problem and working backwards to the model, rather than forwards from a technique.
  • Dive Deep on data and models: Personal, detail-level command of datasets, features, eval results, and error modes in their own work.
  • Deliver Results with models in production: Landing ML systems that run reliably and move business metrics, not just offline wins.
  • Frugality in ML: Treating training and inference cost as a design constraint: right-sizing models, reusing existing assets, cheap experiments first.
  • Bias for Action with scientific honesty: Moving fast on reversible experiments while keeping comparisons and reporting rigorous.

Question themes to expect

  • Working backwards from customer problem to model: Tell me about an ML system you built where you started from a customer problem. How did the customer framing change the technical work?
  • Inference cost versus model quality: Tell me about a time inference cost or latency forced you to change your modeling approach. What did you do?
  • Detail-level command of own model behavior: Pick a model you know best. Tell me about its worst failure mode and how you found it.
  • Rigor and honesty in online experiments: Tell me about an A/B test of your model that gave a result you didn't want. What did you do?
  • Operating models in production: Walk me through an ML production issue you owned end to end: detection, mitigation, root cause, prevention.

Amazon L6 — the loop

  1. Recruiter Screen30 min · Technical recruiter (ML roles)

    Band mapping: this KB's L6 corresponds to Amazon Principal Applied Scientist / Principal MLE (internal L7). Recruiter looks for org-level ML influence: science strategy documents, multi-team model programs, mentorship of senior scientists. Note: no official prep page exists for principal science levels; loop shape extrapolated from the verified senior pattern.

  2. Behavioral Deep-Dive60 min · Principal scientist or principal engineer

    LP rounds at principal calibration: Think Big science bets, Are Right A Lot on modeling directions, Hire and Develop the Best for science teams.

  3. System Design60 min · Principal MLE or applied scientist

    ML platform and portfolio design: shared feature infrastructure, experimentation systems, multi-team model governance, with explicit cost and operational models.

  4. Hiring Manager Round60 min · Director-level hiring manager

    Charter conversation: which org-level ML problems the candidate would own and their thesis. LPs still assigned and probed. Scientist-track loops list a 'tech talk' among evaluated technical competencies — candidates should be ready to present their own research/work to the science community panel; that presentation has no dedicated slot in this KB's round taxonomy and is folded into this deep-dive round.

  5. Leadership / Cross-functional60 min · Bar Raiser, senior, outside the org

    Bar Raiser cross-examines scope authenticity: were the claimed science directions truly the candidate's, and do impact numbers survive detail probing?

What each round scores

  • Think Big in applied science: Setting multi-year ML direction for an org: which capabilities to build, which scientific bets to fund, what to stop.
  • Are Right, A Lot on modeling directions: Judgment on which approaches will transfer to production value, demonstrated over multiple consequential calls.
  • ML platform and governance strategy: Building the shared infrastructure and standards (features, evaluation, deployment gates) that keep many teams' models trustworthy and cheap.
  • Hire and Develop scientific talent: Raising the science bar through hiring and growing senior scientists into principals.
  • Business-grounded science judgment: Allocating science effort by expected business and customer value, including killing interesting work.

Question themes to expect

  • Funding and de-risking a multi-year science bet: Tell me about a science direction you convinced leadership to fund for more than a year before it could pay off. How did you keep it alive and honest?
  • Killing scientifically interesting work: Tell me about a time you shut down an ML effort that the team loved and that was making scientific progress. Why and how?
  • Designing shared ML infrastructure for an org: Design the shared experimentation and deployment infrastructure for an org of eight ML teams with very different latency and risk profiles.
  • Contrarian modeling judgment against the fashion: Tell me about a time the field's momentum pointed one way on a modeling question and you took your org the other way.
  • Connecting model improvements to business value at scale: Walk me through the chain from a model improvement you led to its business impact. I want every link audited.

Amazon L7 — the loop

  1. Recruiter Screen45 min · Senior or executive technical recruiter

    Band mapping: this KB's L7 corresponds to Amazon Senior Principal Applied Scientist / Senior Principal Engineer (internal L8). Screens require company-scale ML evidence: programs spanning business lines, executive advisory, published or widely-cited direction artifacts.

  2. Behavioral Deep-Dive60 min · Senior Principal or Distinguished Scientist/Engineer

    LP coverage at senior-principal calibration: company-scale Think Big, Ownership across business-unit boundaries, Earn Trust with executives.

  3. System Design60 min · Senior Principal or Distinguished Engineer

    Company-scale ML architecture: how a multi-business company should train, share, govern, and pay for models; build/buy/partner judgment on foundational capabilities.

  4. Leadership / Cross-functional60 min · Bar Raiser, very senior, from a distant org

    Bar Raiser scrutiny on authorship and judgment record: separating genuine company-scale influence from proximity to large programs.

  5. Hiring Manager Round60 min · VP or senior director

    Mutual evaluation around a company-level ML charter: the candidate's thesis on where the business's ML leverage actually is.

What each round scores

  • Company-scale ML strategy: Owning ML direction across business lines: capability investment, build/buy/partner on foundational models, and what the company stops doing.
  • Are Right, A Lot on capability timing: Judging when an ML capability is ready to bet a business on — neither early-adopter waste nor late-mover loss.
  • Mechanisms for ML at company scale: Building the standing processes — model governance, evaluation standards, cost review — that keep hundreds of models trustworthy without heroics.
  • Frugality at portfolio scale: Treating company ML compute and inference spend as a portfolio: allocation discipline, efficiency programs, honest unit economics.
  • Executive trust on AI strategy: Being the call executives make before AI commitments: calibrated, candid, and accountable for past advice.

Question themes to expect

  • Owning ML strategy across business lines: Tell me about ML strategy you authored that crossed business-unit boundaries. I want the document, the sponsorship, the resource shifts, and the scoreboard.
  • Build/buy/partner on foundational ML capability: Your company must decide whether to train its own foundation models, fine-tune external ones, or buy via API — at business-line scale. Walk me through how you'd drive this decision.
  • Timing the company's bet on an emerging capability: Tell me about a time you decided when — not whether — your company should bet on an emerging ML capability. How did you judge the timing?
  • Company-wide model governance design: Design the governance that lets a company with hundreds of production models innovate fast while ensuring no model quietly damages customers or the brand.
  • Company compute allocation as a portfolio: You're asked to bring discipline to a nine-figure annual ML compute spend spread across many orgs. Where do you start, and what does the end state look like?

Face the Amazon loop before it faces you

Paste the Amazonjob you're targeting and run a live AI-avatar interview calibrated to this loop — then get a hire/no-hire verdict and a study plan.

Practice free →

Based on publicly reported formats. Not affiliated with or endorsed by Amazon. Loop structures change; verify with your recruiter. Not affiliated with or endorsed by Amazon. Synthesized from public sources; last verified 2026-06-08.

← All companies