
Amazon AI Engineer candidates describe stack-style coding screens, applied ML and LLM questions, and leadership-principle stories. Formats differ between reports, so prepare for both algorithms and ML depth.
$179K
Avg. Base Comp
$247K
Avg. Total Comp
Not reported
Typical Rounds
Not reported
Process Length
Amazon AI Engineer reports pair familiar algorithm screens with applied machine learning questions, and the format appears to shift from one candidate to the next. One candidate was rejected at the online assessment, where the first problem was a stack question in the style of Valid Parentheses. They lost time reaching for a complicated approach before breaking it into smaller cases, and afterward practiced Min Stack and Daily Temperatures. The lesson they took away was to recognize common patterns quickly, not just understand each concept on its own.
A second candidate described a phone screen followed by a virtual loop with two coding rounds, an ML application round, an ML breadth round, and an ML depth round. Each lasted roughly an hour. The coding was LeetCode-style and leaned on CS fundamentals. The ML application round was tied to a real use case for the team, while the breadth and depth rounds were tailored to the candidate's background and touched on CLIP, LLM benchmarks, GRPO, RoPE embeddings, and LLM architecture. Two leadership-principle questions reportedly came up in every round. This candidate also felt the NLP and statistics questions went beyond what the job description and preparation materials suggested.
A third candidate reported an online assessment of multiple-choice aptitude, logic, and grammar questions. It was followed by a technical interview with leadership questions, NLP and LLM fundamentals, and live JSON parsing. Treat that as a sign that assessments differ, not as a standard format.
To prepare, practice timed stack and pattern problems, refresh transformer and LLM fundamentals, write a short project narrative, and have specific leadership stories ready.
Synthesized from 5 candidate 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.
My experience started with an online assessment that was relatively straightforward, followed by a three-round interview loop. Behavioral questions showed up in every round, and the technical bar included medium-to-hard LeetCode problems. The recruiters were genuinely friendly and responsive throughout, which was a refreshing part of the process.
The screening rounds focused on my project experience and the reasoning behind my architectural choices. I was asked to walk through the most technical project I'd done and explain why I'd picked specific approaches. The interviewers wanted to understand my decision-making process, not just whether I knew the right answer. There was also an emphasis on basic generative AI concepts and model names—nothing obscure, but enough to show I keep up with the field.
What bothered me most was that the hiring manager didn't actually conduct the HM interview; someone else stepped in who lacked the depth to really assess my AI/ML background. That round felt like checking boxes rather than having a genuine technical conversation. The overall tone was transactional. I got the sense that they were going through a scripted process and I was one of many candidates being filtered through it. I didn't advance, and frankly, after seeing how mechanical it all felt, I wasn't disappointed. One candidate in a similar timeline did get an offer, though, so there's clearly a path through—it likely just requires aligning perfectly with whatever checklist they're using that cycle.
Prep tip from this candidate
Have a crisp narrative for your most technical project with clear reasoning for each major decision—they drill into 'why' extensively. Brush up on current generative AI model names and capabilities, and practice medium-hard LeetCode regularly. Be aware that the HM interview may not actually involve your hiring manager, so don't expect deep technical vetting at that stage.
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 | |
|---|---|
| Paired Products | |
| 2nd Highest Salary | |
| Upsell Transactions | |
| Longest Streak Users | |
| Rolling Bank Transactions | |
| Customer Orders | |
| Comments Histogram | |
| Top Three Salaries | |
| Closest SAT Scores | |
| Get Top N Frequent Words | |
| Subscription Overlap | |
| Monthly Customer Report | |
| Size of Joins | |
| Random SQL Sample | |
| Merge Sorted Lists | |
| Manager Team Sizes | |
| Download Facts | |
| Top 3 Users | |
| Average Quantity | |
| Month Over Month | |
| 7 Day Streak | |
| Cumulative Reset | |
| Compute Deviation | |
| Always Excited Users | |
| Duplicate Rows | |
| Max Quantity | |
| Flight Records | |
| Cumulative Sales Since Last Restocking | |
| Prime to N |
Synthesized from candidate reports. Individual experiences may vary.
One AI Engineer candidate faced a stack-based coding problem similar to Valid Parentheses and did not pass, mainly because of time lost on an overcomplicated approach. A separate account described multiple-choice aptitude, logic, and grammar questions, so the format may vary.
One AI Engineer candidate reports a phone screening before the main interviews. The report gives little detail about what it covered, so treat it as a first conversation before the technical rounds rather than a defined stage with set content.
That candidate described two LeetCode-style coding rounds, an ML application round tied to the team's use case, and ML breadth and depth rounds tailored to their background. Topics included CLIP, LLM benchmarks, GRPO, RoPE embeddings, and LLM architecture. A separate report mentions live JSON parsing and NLP and LLM fundamentals.
The same candidate reports two leadership-principle questions embedded in every round, which felt repetitive. Another candidate used the STAR method for leadership questions. Prepare concrete stories that can be reused and adapted, since behavioral questions may not be limited to one dedicated round.