← Back to Directory

Amazon interview notes, page 62 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·Technical Phone Screen

May 2026
Amazon ML engineer loop, pretty heavy on the LLM internals side. Whoever designed this round clearly wanted to see if you actually understand the machinery or just know the buzzwords.
  • Walk me through LoRA and its variants like QLoRA, DoRA, LoRA+, and AdaLoRA. How do they differ and when would you pick one over another?
  • Explain RLHF, DPO, and the basics of PPO as they apply to aligning large language models.
  • What is a KV cache, how does it accelerate autoregressive decoding, and what are the memory implications? Also touch on paged attention.
  • Compare different positional encoding schemes: absolute, sinusoidal, learned, RoPE, and ALiBi. What are the tradeoffs?
  • How does FlashAttention achieve speedups over standard attention? Explain the IO-aware design, tiling, and recomputation.

“This is where I spent most of my prep and it showed.”

View Post

Amazon·Software Engineer·Technical Phone Screen

May 2026
Amazon SWE interview with a SQL problem focused on finding users who signed up in a date range but never placed an order during that same window. Pretty standard left join territory but the edge cases around order dates outside the range are where things get interesting.
  • Given a users table and an orders table, write a SQL query to return users who signed up in January 2025 but placed zero orders during that same period. Results should be ordered by user_id ascending, and the solution needs to handle tables with up to 10 million rows each.

“The tricky part isn't the join itself, it's the date scoping on the orders side.”

View Post

Amazon·Data Scientist·Technical Phone Screen

May 2026
Amazon data scientist screen with a classic conditional probability puzzle. Three parts, each hinging on how you interpret the information you're given. The math isn't hard but the reasoning is easy to fumble if you're not careful about assumptions.
  • A family has two children and you know at least one is a boy. What is the probability both children are boys? State your assumptions.
  • Same two-child family, but now you're told specifically that the older child is a boy. Does the probability that both are boys change?
  • Now a child is picked at random from the family and that child happens to be a boy. What is the probability the other child is also a boy?

“I jumped to 1/2 immediately and had to backtrack.”

View Post

Amazon·Software Engineer·Technical Phone Screen

May 2026
Amazon SWE interview with a classic TSP-style problem dressed up in drone delivery language. The n<=15 constraint is a pretty clear hint toward bitmask DP, but if you haven't seen that pattern before you're going to have a bad time.
  • Given a network of delivery hubs and an asymmetric cost matrix, find the minimum fuel cost for a drone to start at hub 0, visit every other hub exactly once, and return to hub 0.

“This is just the Travelling Salesman Problem with a thin coat of drone paint on it.”

View Post

Amazon·Software Engineer·Online Assessment (OA)

May 2026
Amazon SWE online assessment with a pairs-counting problem. Pretty standard OA format, one coding question with constraints that feel straightforward until you think about the edge cases around duplicates.
  • Given an integer array of stock profits (can be negative, zero, or positive) and a target integer, count the number of distinct value pairs (a, b) where a + b equals the target, using two different indices. A value can pair with itself only if it appears at least twice. Return the count of distinct value pairs.

“The duplicate handling is where I almost messed up.”

View Post

Amazon·Data Scientist·Technical Phone Screen

May 2026
Amazon Data Scientist interview with a pandas-heavy analytics question. Pretty focused on practical data wrangling rather than anything algorithmic, which I wasn't totally expecting from a DS round.
  • Given separate users and transactions DataFrames, merge them on user_id keeping only matched rows, then group by country to compute total and average spend, returning the result sorted by total spend descending.

“Knew the mechanics well enough but I second-guessed myself on whether to use an inner join or left join here.”

View Post

Amazon·Software Engineer·Technical Phone Screen

May 2026
Amazon SWE interview with a graph/tree traversal problem about dependency depth. Pretty standard algorithmic question but the dependency framing made me second-guess my approach for a bit.
  • Given a list of nodes where each node has a unique string ID and a list of other nodes it depends on, find the maximum depth of the dependency chain.

“I recognized it as a graph depth problem pretty quickly but the string IDs threw me off at first.”

View Post

Amazon·Software Engineer·Onsite - Behavioral / Leadership

May 2026
Behavioral round at Amazon for an EM role, pretty much one question that took up most of the time.
  • Tell me about a time when a team member was struggling to complete a task.

“Classic leadership question but I fumbled the setup a bit.”

View Post

Amazon·Software Engineer·Online Assessment (OA)

May 2026
Amazon SWE online assessment with a simulation-style coding problem about CPU credit tracking. Pretty straightforward premise but the math tripped me up at first.
  • Given a list of hourly CPU usage rates and a maximum available CPU usage, write a program to calculate the total CPU credits gained or consumed over the time period.

“Took me a minute to realize the credit math is just summing (max_usage minus actual_usage) for each hour, positive means you banked credits, negative means you spent them.”

View Post

Amazon·Technical Program Manager·Onsite - System Design / Architecture

May 2026
Amazon TPM loop with two design questions back to back. Not the most technical grilling I expected but the breadth caught me a bit flat-footed.
  • Design a chess board system.
  • Design a recommendation engine similar to Netflix's.

“I started with the data model, pieces as objects with position and type, and a board as a 2D grid.”

View Post