Amazon Data Scientist interview typically runs 4–7 rounds: online assessment, recruiter screen, technical phone screen, and a 4–5 round virtual onsite loop covering SQL, ML breadth/depth, coding, and a bar raiser. The full process takes roughly 3–6 weeks, with Amazon Leadership Principles woven into every stage.
$123K
Avg. Base Comp
$285K
Avg. Total Comp
4-6
Typical Rounds
2-5 weeks
Process Length
We've coached many candidates through Amazon's Data Scientist loop, and the pattern that separates offers from rejections is almost never pure technical skill. Multiple candidates who felt confident in their ML and coding answers still didn't advance — often because they ran out of strong, end-to-end project stories when interviewers kept drilling for specifics. One rejected candidate put it plainly: they had project examples ready, but not all of them were high-level and fully end-to-end, and that's where they lost it. Amazon's interviewers aren't just checking whether you know what DiD or propensity score matching is — they want to hear you justify a method choice in a real, messy situation you personally navigated.
A recurring theme across recent experiences is that Leadership Principles are woven directly into technical rounds, not siloed into a separate behavioral segment. Candidates reported being asked about Customer Obsession in the middle of an ML project discussion, or being pushed on Disagree and Commit right after a coding question. The Bar Raiser round in particular can feel disorienting because it may contain almost no traditional technical content — one candidate described a final loop where every interviewer focused on leadership questions scoped specifically to data science work, catching them completely off guard after heavy technical prep.
The technical breadth expectation is also genuinely wide. Recent loops have included causal inference without A/B tests, LLM fine-tuning tradeoffs, agentic system design, and SQL window functions — sometimes in the same loop. Candidates who received offers consistently described being comfortable moving quickly between topics rather than going deep on one area. The SQL bar sits around medium difficulty with CTEs, window functions, and joins, and Python data manipulation with pandas is tested practically, not theoretically. If you have a gap anywhere across classical ML, causal inference, or applied statistics, Amazon's loop will find it.
Synthetized from 18 candidates reports by our editorial team.
Had an interview recently?
Share your experience. Unlock the full guide.
Real interview reports from people who went through the Amazon process.
I did not have a recruiter screen. I was asked to fill out a quick questionnaire covering whether I was comfortable coming into the office five days a week, my salary expectations, and whether I had one year of experience with data science fundamentals, SQL, and analysis.
The phone screen was with the hiring manager. It was a deep dive on my resume. There was no coding and no Leadership Principles discussion.
Questions asked:
Share your own interview experience to unlock all reports, or subscribe for full access.
Sourced from candidate reports and verified by our team.
Topics based on recent interview experiences.
Featured question at Amazon
Write a query that returns all neighborhoods that have 0 users.
| Question | |
|---|---|
| 2nd Highest Salary | |
| Rolling Bank Transactions | |
| Customer Orders | |
| Comments Histogram | |
| Closest SAT Scores | |
| Subscription Overlap | |
| Top Three Salaries | |
| Upsell Transactions | |
| Monthly Customer Report | |
| Merge Sorted Lists | |
| Compute Deviation | |
| Experiment Validity | |
| Download Facts | |
| Average Quantity | |
| Random SQL Sample | |
| Manager Team Sizes | |
| Button AB Test | |
| Month Over Month | |
| Flight Records | |
| Prime to N | |
| Paired Products | |
| Swipe Precision | |
| Top 3 Users | |
| Longest Streak Users | |
| Recurring Character | |
| Bank Fraud Model | |
| Jars and Coins | |
| Project Pairs | |
| Always Excited Users |
Synthesized from candidate reports. Individual experiences may vary.
The process typically begins with either a brief recruiter call or a short written questionnaire covering role fit, salary expectations, office attendance requirements, and basic experience with SQL, Python, and data science fundamentals. Some candidates skip a formal recruiter call and go straight to an online assessment or hiring manager screen.
Many candidates complete an online assessment before any live interviews. It typically includes SQL coding questions ranging from basic WHERE filtering to CTEs, RANK, and window functions, plus LeetCode-style Python problems at easy-to-medium difficulty and a leadership principles section testing Amazon-style judgment.
A live screen with a hiring manager or data scientist that combines a resume and project deep dive with technical questions on SQL, Python, and ML fundamentals such as bias-variance tradeoff, regularization, PCA, and model evaluation. Behavioral questions tied to Amazon Leadership Principles, especially Customer Obsession, are often included.
The core of the process is a series of four to six back-to-back virtual rounds covering science depth, science breadth, SQL and data manipulation in Python, a hiring manager case study tied to the team's domain, and a bar raiser behavioral round. Topics span ML theory, deep learning, causal inference, system design, and practical experimentation or evaluation cases, with Leadership Principles woven into nearly every round.
A dedicated bar raiser interview is embedded in the loop and focuses heavily on Amazon Leadership Principles using STAR-format answers. Interviewers probe ownership, conflict resolution, failure cases, and end-to-end project storytelling, often requesting examples scoped specifically to data science or analysis work rather than general project contributions.
One round is typically a team-specific case study led by the hiring manager or a senior scientist. Candidates are asked to clarify an ambiguous problem and propose an analytical approach relevant to the team's work, such as causal inference for business impact estimation, fraud detection from unstructured data, or forecasting without A/B testing.