Amazon Data Scientist Interview Guide

Amazon Data Scientist candidates describe a broad loop rather than a test of one specialty. Reported totals run from two back-to-back rounds to roughly six steps, depending on team and level. Paths commonly open with an online assessment, an HR call or a relaxed hiring manager conversation. A technical screen and a longer loop of several interviews follow. One candidate also reported a reference check after the final round.

SQL shows up repeatedly. One candidate described a hard SQL question in the onsite. Another had SQL folded into a round alongside Leadership Principles. A third was asked to convert rows to columns. Coding ranges from straightforward algorithm problems in an online test to LeetCode-style rounds, with questions such as 2D binary search and computing chamfer distance efficiently.

Machine learning questions mix fundamentals and depth. Reported topics include supervised versus unsupervised methods, the bias-variance tradeoff, regularization, and evaluation metrics like precision, recall and AUC-ROC. One candidate faced a statistics screen that rewarded breadth across the statistical learning toolkit. Another was asked to explain transformer architecture and how KL divergence relates to cross entropy.

Causal inference and open-ended cases deserve real practice time. One candidate was asked to estimate the business impact of a newly launched division. The follow-ups covered when to use difference-in-differences, synthetic control or instrumental variables. Other candidates saw a forecasting prompt and an end-to-end ML pipeline design question.

Leadership Principles appear in most reports, often woven into technical rounds. Prepare specific project stories focused on your data science work, including a disagreement with leadership. Also prepare a clear answer to why you want to work at Amazon.