← Back to Directory

Amazon interview notes, page 138 of 159

FAANG / Big Tech

Most of these Amazon write-ups come from Software Engineers (1,092), then Data Scientists (182) and Product Managers (165). The most reported rounds are the technical phone screen (618), the behavioral / leadership round (309), the online assessment (228) and the system design round (159).

1581 interview notes · updated Oct 2026

Amazon·Machine Learning Engineer·Onsite - System Design / Architecture

Jul 2026
System design round at Amazon for an ML Engineer role, focused entirely on building a distributed data-preprocessing pipeline for LLM training. The question was deep and principal-engineer-level, with follow-ups that kept pushing on every trade-off I made.
  • Design a distributed data-preprocessing pipeline for LLM training. The reference architecture is S3 as the source, feeding into Kafka, then a tokenizer stage, near-duplicate deduplication, a quality filter, and finally an output sink. For each stage, justify your technology choices, explain what breaks at scale (throughput bottlenecks, data skew, fault tolerance, exactly-once semantics, schema evolution), and walk through how you'd monitor and recover from failures.

“This one ran the full hour.”

View Post

Amazon·Software Engineer·Technical Phone Screen

Jul 2026
Amazon SWE coding round, one algorithmic question that sounds deceptively simple but has a bunch of edge cases hiding underneath. Walked through multiple approaches and it turned into a longer conversation than I expected.
  • Given an integer array, find the shortest contiguous subarray that, if sorted in ascending order, would make the whole array non-decreasing. Return the length of that subarray. What are the different approaches and their tradeoffs?

“Started with brute force, check every subarray and verify if sorting it fixes the whole thing.”

View Post

Amazon·Software Engineer·Technical Phone Screen

Jul 2026
Amazon SWE interview with a graph propagation problem that looks straightforward until you realize the adjacency list direction is the whole trap. Solid problem if you've done the LeetCode course schedule stuff, rough if you haven't.
  • Amazon has many internal services with dependencies. Given a set of services being shut down, return all services that become unavailable, including those that depend on them transitively.

“The part that got me was the adjacency list direction.”

View Post

Amazon·Software Engineer·Onsite - System Design / Architecture

Jul 2026
Amazon SWE interview with a product design question that felt more PM-ish than I expected for an engineering role.
  • How would you design the front page of a newspaper app?

“Spent the first few minutes just listing features which was probably the wrong move.”

View Post

Amazon·Machine Learning Engineer·Take-home Assignment

Jul 2026
Amazon ML Engineer interview with a single massive take-home style coding question that basically asked you to build an entire shipment delay prediction system from scratch. The scope was genuinely intimidating and I'm not sure I fully appreciated how many sub-problems were buried inside it until I was already knee-deep.
  • Given a CSV of shipment events with fields like order ID, origin, destination, ship date, promised date, carrier, weight, item count, scan events, and delivered date, build a full Python ML pipeline from scratch that covers data loading and validation, feature engineering, labeling, model training with cross-validation, probability calibration, feature explanation, and finally outputs a CSV of top at-risk shipments with reason codes. The whole thing needs to run in under 5 minutes on 1 million rows using less than 4 GB of memory.

“The runtime and memory constraints are what really got me.”

View Post

Amazon·Software Engineer·Technical Phone Screen

Jul 2026
Amazon SWE coding round, one design-heavy data structure question that started simple and got layered pretty fast. The follow-up about top-X visitors is where things got interesting.
  • Design a system that records user logins and can return the earliest user who has logged in exactly once. Follow-up: extend it to return the top X earliest one-time visitors in order.

“My first instinct was just a hashmap for counts and then iterate everything on each query, which they let me code but then asked about the cost.”

View Post

Amazon·Software Engineer·Onsite - System Design / Architecture

Jul 2026
Did a system design round for a solutions architect role at Amazon. Just one question but it required knowing the actual differences between storage types, not just buzzwords.
  • What is the difference between file storage, object storage, and blob storage?

“I knew the broad strokes but fumbled the blob vs object distinction because they're genuinely close.”

View Post

Amazon·Software Engineer·Technical Phone Screen

Jul 2026
Amazon Research Engineer technical screen, heavy on transformer internals. Felt like a graduate seminar more than a job interview, which I was not fully prepared for.
  • Walk me through the scaled dot-product attention formula, including why you divide by the square root of the key dimension.
  • How does multi-head attention work and what does it gain over single-head attention?
  • Describe the encoder-decoder transformer architecture: what each side is responsible for and how the decoder uses cross-attention over encoder outputs.
  • When would you choose an encoder-only model versus a decoder-only model versus a full encoder-decoder model?
  • Explain padding masks and causal masks in transformers: what each one does and why both are necessary.

“I knew the formula cold but fumbled the intuition behind the scaling.”

View Post

Amazon·Software Engineer·Online Assessment (OA)

Jul 2026
Amazon SWE online assessment with a sliding window / greedy problem that looks approachable until you actually try to pin down the cost accounting. No behavioral, just the one algorithmic problem, but it had enough edge cases to keep me busy.
  • Given a binary array and a fixed window size k, you can repeatedly pick any window of exactly length k, pay the cost equal to the sum of that window, then zero out exactly one 1 inside it. What is the minimum total cost to eliminate all 1s?

“My first instinct was pure greedy: always pick the window with the fewest 1s so the cost per removal is minimized.”

View Post

Amazon·Software Engineer·Onsite - Coding / Algorithms

Jul 2026
Amazon coding round, one meaty grid problem that looked deceptively clean on the surface. The median trick is well-known but the tree-obstacle wrinkle is what they actually care about.
  • Given a set of house coordinates and a set of blocked tree coordinates on a 2D integer grid, find a location for a locker that minimizes the total Manhattan distance to all houses, with the constraint that the locker cannot be placed on a tree cell. Return the minimum total distance and one valid optimal location.

“The median approach clicked pretty fast for me.”

View Post