Company interview guide

Expedia Group Machine Learning Interview Guide (2026)

Recruiter and hiring-manager screens, a modeling assessment or take-home on travel data, then a loop of coding, ML design, ML and stats knowledge and behavioral rounds.

Expedia Group's machine learning interview loop has 8 reported stages, from recruiter call to behavioral & stakeholder round; typical length: 4-10 weeks.

Roles covered: ML Engineer, Applied Scientist and Data Scientist. The question bank holds 6 questions reported from Expedia Group interviews, 1 of them free to solve in your browser.

Try a free question

Interview process length
4-10 weeks
Stages
8
Reports linked to bank questions
6
Bank report dates
Oct 2021–Feb 2026
Questions in bank
6 (1 free)
Last reviewed
October 2026

How the Expedia Group interview loop works

A recruiter call, then either a timed online assessment or a hiring manager screen (often both, in either order), then a final round of three to five interviews of 45-60 minutes that can be spread over several days. ML engineer finals since 2024 most often combine coding, ML system design, ML theory and a behavioral round; some add a Spark or data engineering interview. ML scientist and intern finals usually include a review of your assessment notebook, an ML theory round and a behavioral round. Data science loops mix SQL, statistics and A/B testing, product or stakeholder rounds and, on many teams, a take-home case presented to the panel. Candidates often describe slow scheduling, rescheduled interviews and long waits for a decision, so keep your other processes moving.

The Expedia Group interview process, stage by stage

01Recruiter call

Short phone or video call (one report says 20 min); some candidates had a second recruiter check-in before the final round

A short conversation about your background, why Expedia Group and logistics such as location, level and sponsorship. The recruiter then coordinates the rest of the loop and, according to the company, sends a guide with focus areas when interviews are scheduled. A few candidates found the call rushed, so lead with your most relevant project rather than waiting to be asked.

What it tests

  • Why Expedia Group and why this team
  • A clear two-minute summary of your most relevant ML or data work
  • Logistics: location, relocation, visa sponsorship and timeline

How to prepare

  • Ask which rounds the final will contain and whether an assessment or take-home comes first; loops differ by team.
  • Ask for the interview guide and focus areas if they do not arrive with the scheduling email.
  • Follow up politely if you hear nothing; many candidates report long silences between steps.

A recruiter or HR call is described in 20 data and ML reports, 12 of them since 2024; the company's interview guide says the recruiter coordinates the loop but does not describe a screening call.

02Online assessment

Timed HackerRank-style test, often a notebook modeling task of about 2-2.5 hours to start within a few days; data science versions are about 75 min of SQL, Python and statistics

The most common first filter for machine learning scientist, intern and some data science roles. The classic version hands you a travel dataset and asks you to explore it, engineer features and build a model, for example predicting flight delays, hotel prices or availability, or a multi-class label. Other versions mix SQL, Python and basic statistics, or are proctored multiple-choice tests on ML, data and cloud tooling. If you pass, one final-round interview usually goes through your submission line by line.

What it tests

  • End-to-end tabular modeling under time pressure: cleaning, features, model choice and validation
  • Picking and justifying the most important features
  • SQL and Python on business scenarios, plus basic statistics (p-values, error types, confidence intervals)
  • Defending every modeling choice in a later review of your notebook

How to prepare

  • Rehearse a full notebook workflow on a travel-style dataset within two hours: explore, baseline, improve, evaluate, summarize.
  • Write short markdown notes on why you made each choice; candidates say the review interview asks about exactly that.
  • Some tests are proctored with the camera on and tab switching monitored, so prepare to work without outside help.

A timed online test appears in 26 reports, mostly ML scientist, intern and European data roles, 6 of them since 2024; 10 describe a later round reviewing the submission. The company's guide lists a skills assessment.

03Hiring manager screen

About 30-45 min video call with the hiring manager or team lead; sometimes a director call as well

For most recent ML engineer and data science loops this is the first real interview and the gate to the final round. Expect a mix of resume deep dive, behavioral questions and scenario questions from the manager's own domain, such as how you would validate a model's outputs or check that a KPI is the right one. Some managers ask light technical questions such as simple SQL. One candidate was turned away in minutes because the profile did not match, so tailor your story to the job description.

What it tests

  • A deep dive on past projects: the problem, your decisions and measurable impact
  • Scenario questions on validating models and choosing metrics
  • Fit with the team's domain (search, ads, pricing, recommendations, marketing)
  • Behavioral questions, often in the company's Behaviors framing

How to prepare

  • Prepare one project story with numbers, trade-offs and what you would do differently.
  • Read the job description and the team's engineering blog posts, and map your experience to their problems.
  • Have short answers ready for how you test model outputs and how you approach an unfamiliar dataset.

A hiring manager or team lead interview appears in 22 reports, 14 since 2024, usually right after the recruiter call; in a few older data science loops the manager round came last.

04Take-home case & presentation

Data analysis or modeling case over 24 hours to a week, then a 15-45 min presentation with questions to a panel

Many data science teams, and a few ML teams, send a dataset from the travel business and ask for insights and recommendations: hotel booking nights, ranking hotels from past searches, KPI drivers on hotel stays, or the results of an A/B test. You present the work as slides to two or more interviewers, often inside the final round. One 2025 candidate learned only afterwards that a slide deck was expected, and was told the analysis and model were not deep enough.

What it tests

  • Turning a business dataset into a clear story with recommendations
  • Choosing the right KPI and explaining what drives it
  • A reasonable model with honest validation, not just descriptive charts
  • Handling probing questions on assumptions, data issues and next steps

How to prepare

  • Always build a short slide deck, even if the brief does not ask for one.
  • Structure it as question, data, method, findings, recommendation, risks and extra data you would want.
  • Ask how much time you are expected to spend; some candidates spent ten hours on a vague brief.

A take-home case presented to interviewers appears in 12 data science reports, but only 2 since 2024; the company's interview guide says some roles get a business case or work sample.

05Coding & data round

About 30-60 min live coding in a shared editor or pair-programming tool

The coding interview varies by team. ML engineer and some data science loops report data structure problems (one asked for breadth-first search on a graph), and one 2025 ML engineer loop used a hands-on exercise on data preprocessing and running model inference in Python. ML scientist loops tend toward a live data manipulation or data science task, and data science loops toward SQL with joins, window functions and pivots. Some ML engineer loops add a separate conceptual round on Spark and data engineering.

What it tests

  • Medium data structure and graph problems in Python
  • Data cleaning, preprocessing and model inference code written live
  • SQL: joins, window functions, group by with having, pivots
  • Spark and distributed data processing concepts for ML engineer roles
  • Explaining your thinking while you code

How to prepare

  • Practise pandas and SQL transformations as much as algorithm problems; both appear in recent reports.
  • For ML engineer roles, review Spark basics: partitions, shuffles, joins and when a job is slow.
  • Run your code as you go; one candidate stressed that every solution executed cleanly.

A live coding, SQL or data round appears in 18 reports, 14 since 2024; content splits by track (algorithms or preprocessing for ML engineers, data manipulation for scientists, SQL for data scientists).

06ML system design

About 60 min discussion, usually one round in the final for ML engineer and scientist roles

An open design problem set in Expedia Group's own products. Reported prompts include a two-tower retrieval model for ads, a recommendation system that shows travelers offers, multi-armed bandits for hotel recommendations, what attributes matter for a travel recommender and predicting reservation cancellations. One 2026 candidate found the interviewers unstructured, so drive the conversation with your own outline.

What it tests

  • Retrieval and ranking architecture for recommendations, search and ads
  • Exploration versus exploitation with bandits for hotel or offer ranking
  • Feature and label choices from travel logs (searches, bookings, cancellations)
  • Offline metrics, online experiments and serving constraints

How to prepare

  • Prepare one blueprint for travel recommendations: candidate generation, ranking, features, metrics, A/B test, monitoring.
  • Know the two-tower model in depth, including negative sampling and serving with nearest-neighbor search.
  • Be ready to explain how a bandit would trade off showing new hotels against proven ones.

Free reported questions for this stage

An ML or system design round appears in 17 reports, 11 since 2024, nearly all ML engineer or scientist loops; recommendation and ads problems dominate the questions reported.

07ML & statistics fundamentals

About 45-60 min verbal technical interview; data science loops may split it into a statistics round and a product case

A round of textbook questions plus practical scenarios. ML candidates report bagging versus boosting, supervised versus unsupervised learning, cross validation, overfitting, collinearity, evaluation metrics and which algorithms they would reach for; one ML engineer loop added parallel programming. Data science loops add t-tests, z-tests, A/B test design and product metric cases such as what to measure when a result is not as expected.

What it tests

  • Tree ensembles, bias and variance, regularization and model evaluation
  • Hypothesis tests and checking whether two samples come from the same distribution
  • A/B test design, power and interpreting unexpected results
  • Product and business metrics for a travel marketplace

How to prepare

  • Be able to explain bagging versus boosting crisply; it is the most repeated single question in our reports.
  • Review the statistics behind A/B tests: p-values, error types, confidence intervals and sample size.
  • Tie each answer to a travel example, such as booking conversion or hotel ranking.

ML theory, statistics or A/B testing questions appear in 27 reports, 11 since 2024; ML roles get ML theory, data science roles get statistics and product cases.

08Behavioral & stakeholder round

About 45-60 min, sometimes with a senior manager or product partner

Nearly every final round has one behavioral interview, and data science loops often add a stakeholder conversation with a product or senior manager. The company says it uses questions about past experience answered in STAR form and assesses fit with its five Behaviors. Reported questions cover ambiguous requirements, conflict with colleagues, mistakes, teaching yourself a skill and managing priorities.

What it tests

  • Handling ambiguous requirements and shifting priorities
  • Working with stakeholders and explaining technical results simply
  • Conflict, mistakes and what you learned
  • Motivation for Expedia Group and the travel domain

How to prepare

  • Prepare six to eight STAR stories and map each to the company's Behaviors, such as Traveler First and Ownership Mindset.
  • Include one story about explaining a model to a non-technical stakeholder.
  • Have specific questions ready for the interviewer about how the team measures success.

A behavioral or stakeholder interview appears in 31 reports, 13 since 2024; the company's interview guide describes STAR-style behavioral questions and an assessment against its Behaviors.

Questions and themes reported most often

  • Present a take-home analysis of hotel or A/B test data to a panel (12 reports, take-home case & presentation)
  • Build a model on a dataset, often travel data, in a timed notebook (flight delays, hotel prices or availability) (10 reports, online assessment)
  • Walk the interviewer through your assessment submission and defend each choice (10 reports, online assessment)
  • A/B test design and analysis: test versus control, metrics, unexpected results (5 reports, ml & statistics fundamentals)
  • SQL with joins, window functions and group by with having (5 reports, coding & data round)
  • Design a recommendation or ads system for travelers, including two-tower retrieval (4 reports, ml system design)
  • Spark and data engineering concepts for ML engineers (4 reports, coding & data round)
  • Bagging versus boosting (3 reports, ml & statistics fundamentals)

What Expedia Group asks about

  • Behavioral, hiring manager & project deep dive22%
  • Modeling assessment & take-home cases22%
  • ML & statistics fundamentals20%
  • Coding, SQL & data engineering19%
  • ML system design17%

Expedia Group interview prep checklist

  • Built and documented a full model on a hotel or flight dataset in under two and a half hours
  • Can explain every feature and modeling choice in that notebook out loud
  • Sketched a travel recommender with retrieval, ranking, metrics and an A/B test plan
  • Can explain bagging versus boosting and a two-sample test in under two minutes each
  • Solved SQL problems with window functions and pivots, and a few medium graph problems
  • Prepared a slide-deck template for a take-home case: question, method, findings, recommendation, risks
  • Have six STAR stories mapped to the company's Behaviors

Tips

  • Ask the recruiter early which track you are on: assessment first, hiring manager first, or a take-home case; the preparation differs.
  • Practise a two-hour notebook on a travel dataset until you can explore, model, validate and summarize without rushing.
  • Annotate any assessment or take-home thoroughly; a later interview usually goes through it in detail.
  • Prepare a travel recommendation and ranking design, including two-tower retrieval and bandit exploration.
  • Drill bagging versus boosting, overfitting, cross validation and evaluation metrics until the answers are short and exact.
  • For data science roles, review SQL window functions and the statistics behind A/B tests.
  • Prepare STAR stories tied to the company's Behaviors and at least one about a stakeholder disagreement.
  • Expect slow scheduling and rescheduled interviews, and keep other processes moving in parallel.

See the rest of the Expedia Group loop

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

Unlock with Pro

Practice the Expedia Group loop

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

Expedia Group interview FAQ

How many interview rounds does Expedia Group have for ML roles?

Candidates report 8 stages: Recruiter call, Online assessment, Hiring manager screen, Take-home case & presentation, Coding & data round, ML system design, ML & statistics fundamentals and Behavioral & stakeholder round. A recruiter call, then either a timed online assessment or a hiring manager screen (often both, in either order), then a final round of three to five interviews of 45-60 minutes that can be spread over several days. ML engineer finals since 2024 most often combine coding, ML system design, ML theory and a behavioral round; some add a Spark or data engineering interview. ML scientist and intern finals usually include a review of your assessment notebook, an ML theory round and a behavioral round. Data science loops mix SQL, statistics and A/B testing, product or stakeholder rounds and, on many teams, a take-home case presented to the panel. Candidates often describe slow scheduling, rescheduled interviews and long waits for a decision, so keep your other processes moving.

How long does the Expedia Group ML interview process take?

Reported timelines run 4-10 weeks.

What does Expedia Group ask in ML interviews?

By reported question mix: Behavioral, hiring manager & project deep dive 22%; Modeling assessment & take-home cases 22%; ML & statistics fundamentals 20%; Coding, SQL & data engineering 19%; ML system design 17%. The most repeated themes: Present a take-home analysis of hotel or A/B test data to a panel; Build a model on a dataset, often travel data, in a timed notebook (flight delays, hotel prices or availability); Walk the interviewer through your assessment submission and defend each choice.

How should I prepare for the Expedia Group ML interview?

Ask the recruiter early which track you are on: assessment first, hiring manager first, or a take-home case; the preparation differs. Practise a two-hour notebook on a travel dataset until you can explore, model, validate and summarize without rushing. Annotate any assessment or take-home thoroughly; a later interview usually goes through it in detail.

Are there free Expedia Group interview questions?

Yes. LeetML has 6 questions reported from Expedia Group interviews, and 1 of them are free to solve in the browser; they are listed by stage above.

Related interview guides

Official sources