American Express Data Scientist Interview Guide

What stands out most about the American Express Data Scientist process is how deliberately it blends domain knowledge with technical fundamentals. This isn't a pure LeetCode shop. Both candidates we heard from noted that the SQL and coding portions felt more like a baseline check than a serious filter — what actually separated the conversations was depth on past projects, particularly anything touching credit risk or financial modeling. If you've worked on a credit scoring or risk model, expect to defend every decision you made, from feature selection to how you handled class imbalance.

A recurring theme across both experiences is that AmEx interviewers probe practical judgment, not just textbook answers. The question about handling a variable with 80%+ missing values, or how to avoid overfitting when data is sparse for a given category — these aren't gotcha questions, but they do require you to think out loud about real tradeoffs. We've seen candidates stumble here by jumping to a technique without explaining why it fits the problem. The ensemble model questions (XGBoost vs. Random Forest, Lasso vs. Ridge) follow the same pattern: they want reasoning, not recitation.

The behavioral component is taken more seriously here than at many comparable firms. Multiple candidates flagged that communication style and culture fit felt like genuine evaluation criteria, not just box-checking. Being concise but thorough in how you explain your thinking — whether on a statistics question or a conflict with a teammate — appears to matter as much as getting the answer right. The Gurgaon-based senior manager round also signals that this role sits within a larger global team, so demonstrating awareness of the financial services context, especially the Indian credit market if relevant, can make a real difference.