Amazon Data Engineer Interview Loop
The Amazon data engineer interview follows Amazon's usual process. There's a recruiter screen, then either a 90-minute timed online assessment or a technical phone screen, depending on the team. Then comes a loop of four to five rounds. Two are deep technical, two are Leadership Principles, and one is the Bar Raiser, an outsider to the team whose job is to say no. Amazon is the one company here where the behavioral rounds fail as many candidates as the technical ones.
What decides it
Leadership Principles, tested in every round. A technical interviewer will ask about a time you disagreed with a decision, and they'll listen for specifics and numbers. Vague stories are the most common reason strong engineers get rejected here. Prepare six to eight real stories and match each one to a principle. The Bar Raiser will ask the hardest version.
Practice by round
SQL and Python basics under time, in whichever format your team uses. Expect orders, shipments and employees, plus word counts and common elements in Python. Some candidates say they couldn't move on with a single failing test case.
The two deep-technical rounds are where this role is really assessed. The SQL is on orders, shipments and customers: rolling monthly sales, month-over-month change, customers without an order, and the percentage of shipable orders. The Python looks like pipeline work: flattening nested JSON, intersecting arrays, a linked-list stack. The most common reason for rejection is ignoring scale. Say what breaks first and what you'd do about it, before anyone asks.
Design reasoning on Amazon's own problems, like a database for customer addresses, a replacements-recommendation system, or when a linked list beats an array. Then come the mini-analyses and projects, the end-to-end version: resolving duplicate transactions, delivery promise misses, split shipment rate and processing image files. Some teams fold design into the technical rounds.
| Case study | Difficulty |
|---|---|
| Database for Customer Address | Medium |
| Replacements Recommendation | Medium |
| Slowly Changing Seller Dimension | Medium |
| Order Header or Order Line: Which Grain? | Medium |
| Canary and Rollback for a Ranking Model | Medium |
| Data project | Difficulty |
|---|---|
| Buy Again Rollout | Hard |