Company interview guide

American Express Machine Learning Interview Guide (2026)

Tree-based ML fundamentals, SQL, business cases and guesstimates, with heavy weight on your own projects.

American Express's machine learning interview loop has 6 reported stages (Recruiter or HR call (lateral hires), Campus test (campus hires), Technical interview: ML, statistics & your projects, SQL & Python coding, Case study, guesstimates & puzzles, Hiring manager, senior leaders & behavioral); typical length: 1 day (campus) to 2+ months (lateral).

Roles covered: Data Scientist, ML Engineer and AI Engineer. The question bank holds 2 questions reported from American Express interviews, 1 of them free to solve in your browser.

Try a free question

Interview process length
1 day (campus) to 2+ months (lateral)
Stages
6
Reports linked to bank questions
2
Bank report dates
Oct 2022–Aug 2024
Questions in bank
2 (1 free)
Last reviewed
October 2026

How the American Express interview loop works

Two common shapes. Campus hiring, mostly in India: a written or online test (aptitude plus an ML or business-case section in one report), then two elimination interviews, a technical and case round with a manager or senior analyst and a final round with a VP or hiring manager, often on the same day. Lateral hiring, mostly in the US: a short recruiter or HR call, a technical video interview on ML, statistics, SQL and your projects, then a final stage of two to five conversations that mixes a case study, puzzles, behavioral questions and a hiring manager, director or VP. One senior ML engineer loop instead ran recruiter, two coding questions, ML concepts, then a paper-reading final; one older loop had a one-hour presentation. Several candidates say the recruiter's description of a round did not match what was asked, so ask for specifics. The company's guidance bans AI tools in interviews unless the interviewer allows them.

The American Express interview process, stage by stage

01Recruiter or HR call (lateral hires)

About 10-30 min by phone with a recruiter or HR

A quick fit check on background, the role, relocation and salary. Some recruiters also ask how strong you are at machine learning and coding. Campus candidates usually skip this and go straight to a test.

What it tests

  • Whether your background matches the role and team
  • Logistics: relocation, salary expectations, timeline
  • A short self-rating of your ML and coding experience

How to prepare

  • Ask what each later round will cover; more than one candidate prepared for the wrong format because the description was vague or wrong.
  • Have a one-minute summary of your most relevant modelling project, ideally in risk, fraud, credit or marketing.

A recruiter or HR phone call opens 10 lateral-hire reports from 2019-2026 (four since 2024), described as 10-30 minutes on background, logistics and salary; campus reports skip it.

02Campus test (campus hires)

Written or online test on campus before interviews; one report describes aptitude plus a choice of an ML section or a business-case section

Campus drives shortlist on resume, then run a test that decides who gets interviews. One candidate faced aptitude questions followed by a choice between a basic ML and data science section and a business case section; another took an on-campus coding test with the whole candidate pool. The company's student pages describe pre-recorded video and technical assessments for some campus tracks.

What it tests

  • Quantitative aptitude and logical reasoning
  • Basic ML and data science concepts
  • A short business case, as an alternative section
  • Basic coding logic

How to prepare

  • Practise timed aptitude and probability questions; they gate the interviews.
  • If you get a choice of sections, pick the one where you can show depth; the interviews that follow probe it.

A campus test before interviews in 3 reports (2019, 2022, 2025). The company's student programs page lists video interviews and technical or coding assessments for several campus tracks.

03Technical interview: ML, statistics & your projects

Usually one video or in-person interview of up to an hour; on campus, the first of two elimination rounds

The core round. Expect a walk through the projects on your resume with pointed follow-ups, then classical ML questions, most often on decision trees, random forests, bagging versus boosting and XGBoost, plus overfitting, imbalanced classes, missing data and model interpretation. Campus rounds add statistics and probability such as p-values and the central limit theorem. Some interviewers ask about the actual project the team is hiring for. Neural network and NLP questions show up for ML and AI engineer roles.

What it tests

  • Random forest versus gradient-boosted trees, bagging versus boosting, XGBoost parameters, Gini and splits
  • Overfitting, regularization and handling small or imbalanced classes
  • Missing values, feature handling and model interpretation
  • Statistics basics: p-values in plain words, the central limit theorem, linear regression
  • Depth on your own projects: why each choice, how you tuned it, what it achieved

How to prepare

  • Be able to explain trees end to end: split criteria, why a forest reduces variance, how boosting differs and which XGBoost parameters control overfitting.
  • Prepare to defend every number and method on your resume; several candidates were grilled on their projects for most of the round.
  • Practise explaining a p-value and the central limit theorem to a non-technical listener.
  • Link your answers to credit risk or fraud where you can; interviewers often frame questions that way.

ML, statistics or project questions in a technical round in 17 reports (12 since 2024), mostly data scientist; tree ensembles come up in eight. Consistent across campus and lateral loops.

04SQL & Python coding

Usually part of the technical interview; some loops add a separate live coding session (two coding questions in one ML engineer loop, a live session with teammates in another)

Data scientists mostly get SQL questions and basic Python, sometimes analysing a dataset in Python or discussing how to optimize a query. ML and AI engineer loops are more likely to have a dedicated coding round with standard algorithm questions; one asked to merge several sorted linked lists. Reports describe the coding as easy to moderate.

What it tests

  • Writing SQL queries and explaining how you would speed them up
  • Basic Python and exploring a dataset
  • Standard data-structure questions for engineering roles (heaps, linked lists)
  • Explaining code logic clearly, sometimes by hand

How to prepare

  • Brush up on joins, window functions and aggregation; SQL is the most common coding topic for data scientists.
  • For engineering roles, practise common heap and linked-list problems under time pressure.
  • Do not use AI tools during interviews; the company's candidate guidance forbids it unless the interviewer allows it.

Free reported questions for this stage

SQL, Python or live coding in 14 reports (8 since 2024); a separate coding round in 4, otherwise inside the technical interview. The student programs page lists coding assessments for some tracks.

05Case study, guesstimates & puzzles

A case or resume round on campus, or one of the final-stage interviews for lateral hires; length not reported

Business reasoning carries real weight. Cases are usually card-business problems: how you would detect credit card fraud, which metrics matter for a new card, or a marketing data science case. Campus interviews, particularly in India, mix in guesstimates and logic puzzles, and several candidates were asked how American Express makes money. One lateral candidate said the case did not match what the recruiter had described.

What it tests

  • Framing a fraud, credit or marketing problem as a model: target, features, metrics, imbalance
  • Market-sizing guesstimates with clearly stated assumptions
  • Logic puzzles and brain teasers
  • Knowing the card business model and the credit market

How to prepare

  • Learn how a card issuer earns money (discount revenue, card fees, interest) and how credit risk and fraud losses affect it.
  • Practise a fraud detection case end to end: label definition, features from transactions, class imbalance, precision versus recall costs.
  • For guesstimates, say every assumption out loud and sanity-check the final number.

A case study, guesstimate or puzzle in 14 reports (10 since 2024), most from data scientist campus loops; fraud or card-product cases appear in five and the business model question in three.

06Hiring manager, senior leaders & behavioral

On campus a final round with a VP or hiring manager; for lateral hires two to five conversations of about 30-60 min with the hiring manager, team members, directors or a VP, sometimes on the same day as the technical round

The last stage checks problem solving, business sense, communication and fit. Some loops are a single manager interview; others are a panel or a long series of calls with directors focused on leadership and working across teams. Expect classic behavioral prompts on conflict, coordination and a challenging project, and why you want to join. Rarer formats include a one-hour presentation with technical questions and, for one senior ML engineer, discussing a research paper.

What it tests

  • Behavioral stories: conflict with a teammate, coordinating with another team, a hard project
  • Business understanding and stakeholder communication
  • Motivation: why American Express and how you would apply data science there
  • Leadership and collaboration for senior roles

How to prepare

  • Prepare three or four stories with a clear result, including one about getting data or help from another team.
  • Have a specific answer to how your skills would help in risk, fraud or marketing at the company.
  • Ask whether the final stage includes a presentation or paper discussion so you can prepare for it.

A hiring-manager, VP, director or behavioral round in 14 reports (8 since 2024); format ranges from one interview to six calls. The student programs page describes behavioral and competency interviews.

Questions and themes reported most often

  • Deep dive on the projects listed on your resume (11 reports, technical interview: ml, statistics & your projects)
  • Random forest vs gradient boosting, bagging, XGBoost parameters (8 reports, technical interview: ml, statistics & your projects)
  • Guesstimate or market-sizing question (6 reports, case study, guesstimates & puzzles)
  • Logic puzzle or brain teaser (6 reports, case study, guesstimates & puzzles)
  • SQL questions and query optimization (5 reports, sql & python coding)
  • Statistics basics: p-value in plain words, central limit theorem, probability (4 reports, technical interview: ml, statistics & your projects)
  • Explain the American Express business model or the credit card market (4 reports, case study, guesstimates & puzzles)
  • Conflict, cross-team coordination and a challenging project (4 reports, hiring manager, senior leaders & behavioral)

What American Express asks about

  • ML fundamentals (trees, boosting, overfitting)28%
  • Business case, guesstimates & puzzles22%
  • Resume projects & behavioral20%
  • SQL & Python coding17%
  • Statistics & probability13%

American Express interview prep checklist

  • Can explain random forest vs gradient boosting and name the XGBoost parameters that fight overfitting
  • Rehearsed a five-minute deep dive on each resume project, with numbers
  • Walked through a credit card fraud model: labels, features, imbalance, thresholds
  • Can explain how American Express earns money
  • Solved a few guesstimates and logic puzzles out loud
  • Practised SQL joins, window functions and aggregations
  • Prepared behavioral stories on conflict and cross-team work

Tips

  • Make tree ensembles your strongest topic: splits, random forests, boosting, XGBoost tuning and interpretation come up more than anything else.
  • Rehearse every project on your resume in depth, including the business impact and why you chose each method.
  • Prepare a credit card fraud modelling case end to end, including how class imbalance changes your metric choice.
  • Learn the basics of how a card issuer makes money; interviewers ask it directly and use it to frame cases.
  • For campus interviews, practise guesstimates and logic puzzles out loud alongside probability and statistics.
  • Keep SQL sharp (joins, window functions, aggregation) and be ready to talk about query performance.
  • Ask the recruiter what each round covers, and confirm again before the final stage; descriptions have been inaccurate.
  • Do not use AI tools or assistive devices in interviews unless the interviewer explicitly allows it; the company says misuse can disqualify you.

See the rest of the American Express loop

The later stages, the questions reported for them and how to prepare are in Pro.

Unlock with Pro

Practice the American Express loop

Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.

American Express interview FAQ

How many interview rounds does American Express have for ML roles?

Candidates report 6 stages: Recruiter or HR call (lateral hires), Campus test (campus hires), Technical interview: ML, statistics & your projects, SQL & Python coding, Case study, guesstimates & puzzles and Hiring manager, senior leaders & behavioral. Two common shapes. Campus hiring, mostly in India: a written or online test (aptitude plus an ML or business-case section in one report), then two elimination interviews, a technical and case round with a manager or senior analyst and a final round with a VP or hiring manager, often on the same day. Lateral hiring, mostly in the US: a short recruiter or HR call, a technical video interview on ML, statistics, SQL and your projects, then a final stage of two to five conversations that mixes a case study, puzzles, behavioral questions and a hiring manager, director or VP. One senior ML engineer loop instead ran recruiter, two coding questions, ML concepts, then a paper-reading final; one older loop had a one-hour presentation. Several candidates say the recruiter's description of a round did not match what was asked, so ask for specifics. The company's guidance bans AI tools in interviews unless the interviewer allows them.

How long does the American Express ML interview process take?

Reported timelines run 1 day (campus) to 2+ months (lateral).

What does American Express ask in ML interviews?

By reported question mix: ML fundamentals (trees, boosting, overfitting) 28%; Business case, guesstimates & puzzles 22%; Resume projects & behavioral 20%; SQL & Python coding 17%; Statistics & probability 13%. The most repeated themes: Deep dive on the projects listed on your resume; Random forest vs gradient boosting, bagging, XGBoost parameters; Guesstimate or market-sizing question.

How should I prepare for the American Express ML interview?

Make tree ensembles your strongest topic: splits, random forests, boosting, XGBoost tuning and interpretation come up more than anything else. Rehearse every project on your resume in depth, including the business impact and why you chose each method. Prepare a credit card fraud modelling case end to end, including how class imbalance changes your metric choice.

Are there free American Express interview questions?

Yes. LeetML has 2 questions reported from American Express interviews, and 1 of them are free to solve in the browser; they are listed by stage above.

Related interview guides

Official sources