Amazon ML Engineer interview typically runs 7 rounds: online assessment, HR call, coding interview, 2 coding interviews, 2 system design interviews, and a behavioral interview. It usually takes about 2 weeks to complete and is notably assessment-heavy.
$135K
Avg. Base Comp
$198K
Avg. Total Comp
4-5
Typical Rounds
3-5 weeks
Process Length
Our candidates report that Amazon’s ML Engineer process is less about flashy model theory and more about whether you can turn a product problem into a scalable, defensible system. The questions skew toward recommender systems, ranking, retrieval, and ad matching, which tells us the team is looking for people who can reason about tradeoffs in real production settings rather than just name the right algorithm. Even the more conceptual prompts, like bias-variance, class imbalance, or SVMs versus deep learning, were framed in a way that pushed candidates to connect the answer back to business impact and system behavior.
A recurring theme is that the bar rises sharply once the conversation moves into ML system design. Multiple candidates described being asked to design ML algorithms or ranking systems while also thinking through complexity constraints, including getting pieces of the solution to O(log n). That combination is revealing: Amazon seems to care a lot about algorithmic efficiency inside ML workflows, not just whether the model is sound. We also saw one candidate note that the online assessment felt like the main filter, which suggests the company uses early technical signal to screen for persistence and precision before investing in deeper conversations.
What makes this process feel distinctly Amazon is the emphasis on practical rigor over polish. The interview set included straightforward coding problems alongside harder ML design prompts, so candidates who only prepared one side tended to feel exposed. The strongest signal, based on these experiences, is the ability to move fluidly from code to system to product consequence without losing clarity. That’s the pattern we’d coach around here: Amazon wants ML engineers who can build something that works, scales, and fits a customer-facing use case.
Synthetized from 1 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.
The hardest part for me was the online assessment. After I applied, Amazon sent me a HackerRank-style assignment with two LeetCode questions, and the second one was especially tough. I couldn’t finish it, and even after the fact I wasn’t able to track down a solution online, which was frustrating because that round felt like the main filter. I also had a referral, but I never spoke with the hiring manager at any point in that process.
The rest of the loop was more standard on paper but still pretty demanding. After an HR call, I did a one-hour coding interview with a team member, and then there was a five-round loop made up of two LeetCode-style coding interviews, two system design interviews, and one behavioral interview. The coding questions were described as easy to medium in difficulty, but the system design rounds were more ML-focused and asked me to design ML algorithms or a ranking system, including thinking through time complexity and getting parts of the solution down to O(log n). I also had a separate process where a recruiter reached out because I had applied for another role, and that path moved quickly over about two weeks with one recruiter interview and one hiring manager conversation. Overall, the assessment and the ML/system design depth were the main things to prepare for, not just generic coding practice.
Prep tip from this candidate
Practice Amazon-style HackerRank screens with two timed coding problems, including at least one harder question that may not be straightforward to brute force. For ML system design, be ready to talk through a ranking system and justify time-complexity choices, including how you’d get key operations to O(log n).
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
Given two sorted lists, write a function to merge them into one sorted list.
| Question | |
|---|---|
| Compute Deviation | |
| Permutation Palindrome | |
| Bagging vs Boosting | |
| P-value to a Layman | |
| Get Top N Frequent Words | |
| Prime to N | |
| Compute Variance | |
| Recurring Character | |
| Bank Fraud Model | |
| Random Forest Explanation | |
| Weekly Aggregation | |
| Jars and Coins | |
| Type-ahead Search | |
| Encoding Categorical Features | |
| Same Algorithm Different Success | |
| Flatten JSON | |
| Valid Anagram | |
| Hurdles In Data Projects | |
| Booking Regression | |
| Lasso vs Ridge | |
| Biased Random Number Generator | |
| Duplicate Rows | |
| Assumptions of Linear Regression | |
| Perfectly Separable | |
| Dice Worth Rolling | |
| Move Zeros Back | |
| N Dice | |
| Impute Median | |
| Matrix Rotation |
Synthesized from candidate reports. Individual experiences may vary.
After applying, candidates receive an online assessment with two LeetCode-style coding problems. In this experience, the second question was especially difficult and appeared to be the main filter before moving forward.
An HR recruiter call follows the assessment for candidates who pass the initial screen. This stage is a standard introductory conversation before the technical interviews.
Candidates complete a one-hour coding interview with a team member. The questions are described as easy to medium difficulty and are similar to LeetCode-style problems.
The main loop consists of two coding interviews, two system design interviews, and one behavioral interview. The system design rounds are ML-focused and may ask candidates to design ML algorithms or a ranking system, with attention to time complexity and optimizing parts of the solution to O(log n).