Company interview guide
Indeed Machine Learning Interview Guide (2026)
A take-home modeling task, then rounds on coding, probability and statistics, and ML that stay close to classic data science.
Indeed's machine learning interview loop has 8 reported stages, from recruiter call to hiring manager and wrap-up; typical length: 3 weeks to 4 months.
Roles covered: Data Scientist and ML Engineer. The question bank holds 9 questions reported from Indeed interviews, 2 of them free to solve in your browser.
- Interview process length
- 3 weeks to 4 months
- Stages
- 8
- Reports linked to bank questions
- 9
- Bank report dates
- Feb 2020–Sep 2026
- Questions in bank
- 9 (2 free)
- Last reviewed
- October 2026
How the Indeed interview loop works
A recruiter screen, then usually a take-home modeling assignment (often predicting salaries from a job dataset) or, in some offices, a timed online test mixing probability, coding and a small ML task. A one-hour technical screen with two data scientists follows, covering easy coding, probability, statistics and ML basics. The final stage is a set of about four focused rounds, each typically with two interviewers: coding, math and statistics, machine learning, and a 60 to 90 minute solo data exercise, closing with a short conversation with a manager or senior leader. The company says interviews are now virtual and that multi-round decisions can take several weeks; candidates report anything from three weeks to four months. Machine learning engineer reports from 2026 describe an ML design phone screen and ML coding from scratch, with a harder data-structure coding round later. This outline rests on about two dozen reports, most of them from 2021-2022 and earlier.
The Indeed interview process, stage by stage
01Recruiter call
Short phone or video call with a recruiter; some last only a few minutes
A conversation about your background, why you are looking and why Indeed, plus an overview of the stages ahead. Work authorization can end the process here if the role does not sponsor visas. A few candidates were sent the take-home before speaking to anyone.
What it tests
- Why Indeed and why now
- A quick walk through your resume and recent projects
- Logistics: location, sponsorship, timeline
How to prepare
- Ask the recruiter to list every round and its focus; the process is structured and recruiters can usually describe it.
- Raise sponsorship needs early; at least one candidate was stopped on that alone.
- Have a two-minute summary of a project that used data to change a product decision.
Recruiter or HR screen opens the loop in 11 data science reports (3 since 2024); the company's hiring page lists a recruiter conversation first.
02Take-home or online test
Take-home dataset task with about a week to submit; some offices use a timed online test instead
Most reports describe a take-home: analyze a dataset and build a predictive model, often predicting salaries from job-posting features, then answer a set of written questions about your choices and error. Several candidates found it took longer than the suggested time. In some offices a timed online test (probability, a small ML task on provided data, and basic coding) replaces or adds to it. Some teams ask you to present the work to a panel; others never mention it again.
What it tests
- Cleaning mixed categorical and numerical features
- Choosing, validating and comparing regression models
- Explaining modeling choices and error rates in writing
- Basic probability and coding under time pressure (online test)
How to prepare
- Treat the written answers as part of the grade: justify every choice and report error on held-out data.
- Start with a simple baseline, then show what each added feature or model buys.
- Budget more time than suggested; several candidates said it ran long.
- Be ready to present and defend the work, in case your team uses a panel review.
Take-home in 14 data science reports (2015-2026, incl. 2024 and 2026) and a timed online test in 3; the company's interview tips page mentions take-home or coding assessments.
03Technical screen
About 60 min by video with two data scientists; shared editor for live coding
A rapid-fire round across coding, probability, statistics and ML basics, typically one interviewer leading and a second shadowing. Coding is easy to medium (list manipulation, sorting, a product-of-others array), probability is textbook (draws with and without replacement, Bayes rule), and statistics covers hypothesis tests and A/B testing. In 2025-2026 candidates also had to write gradient descent or linear regression from scratch, and a 2026 machine learning engineer screen was a full ML design discussion instead.
What it tests
- Easy to medium Python coding with complexity analysis
- Counting probability with and without replacement
- Hypothesis testing, t-test vs z-test, A/B test basics
- ML basics: missing data, overfitting, loss functions
- Implementing a simple model from scratch in numpy
How to prepare
- Expect many short questions in one hour; answer crisply and move on.
- State the complexity of every function you write, including built-in sorts.
- Practise writing linear regression with gradient descent and MSE without libraries.
- Drill probability set-ups until you can write the expression before computing it.
Free reported questions for this stage
- Linear Regression Using Gradient DescentML CodingMediumSolve
One-hour screen with coding plus probability, statistics or ML in 14 reports (5 since 2024, two from 2026 ML engineer screens); two interviewers mentioned in four.
04Coding round
About 50-60 min live or whiteboard-style coding, often with two engineers
Data scientist reports describe easy to medium problems discussed at a whiteboard, sometimes with data wrangling or SQL, and follow-up variations that add constraints. A 2026 machine learning engineer onsite was harder: implementing an unrolled linked list with a specific node-splitting rule.
What it tests
- Clean code for list and string problems with follow-up variations
- Data wrangling with dataframes or SQL joins
- Talking through complexity as the problem changes
- Data-structure implementation (ML engineer track)
How to prepare
- Narrate your approach before coding; interviewers grade the reasoning as well as the result.
- Practise follow-ups such as removing duplicates or handling streams after a basic merge.
- For ML engineer roles, practise implementing a small custom data structure end to end.
Separate coding or data wrangling round in 6 data science reports (2015-2022) and one 2026 ML engineer onsite; the company's hiring page notes whiteboard or technical exercises.
05Math, statistics and probability
About 50-60 min, usually two interviewers
A dedicated round on Bayes rule, conditional probability, hypothesis testing, regression and experiment design. Questions are classic and well defined, but interviewers push through several follow-ups, such as testing for normality or comparing non-normal samples.
What it tests
- Bayes rule with coins or mixed populations
- Choosing and interpreting a hypothesis test
- Linear regression assumptions and interpretation
- Designing and reading an A/B test
How to prepare
- Review non-parametric tests and when a t-test is the wrong tool.
- Write out the sample space before any probability calculation.
- Connect each test to a job-marketplace example, such as an apply-rate experiment.
Dedicated math and statistics round in 5 final-stage reports, all 2015-2022; probability and statistics also appear in most technical screens.
06Machine learning and ML design
About 50-60 min, discussion with two interviewers; ML system design for senior and ML engineer roles
Data scientist rounds cover how models learn, objective functions, hyperparameter tuning, over- and underfitting, and sometimes NLP. Senior and ML engineer candidates since 2025 report full ML system design around Indeed's own products: a job recommendation system, or rebuilding a job-spam classifier that scores well offline but misses in production, including rollout, online testing and drift monitoring.
What it tests
- Objective functions, regularization and the bias-variance trade-off
- Model evaluation and hyperparameter tuning
- End-to-end design of a job recommendation or job-quality model
- Diagnosing offline versus online gaps, rollout and A/B testing
- Cost and latency trade-offs between tree models, neural nets and small language models
How to prepare
- Prepare one design each for job recommendations and spam or low-quality job detection.
- Cover data, labels, features, loss, offline metrics, online test and monitoring in every design.
- Expect questions about LLM-generated or adversarial content in classification problems.
Free reported questions for this stage
- Rebuilding a Job-Post Spam Classifier That Fails in ProductionML System DesignHardSolve
ML round in 5 data science final-stage reports (2015-2022) and a job recommendation design in one 2025 senior report; a 2026 ML engineer screen was ML design too.
07Data exercise
About 80-90 min working alone on a data task, then discussing it
A timed, hands-on data science exercise relevant to the business, done quietly on your own; one candidate's was about recommendation systems and called it the hardest round. Treat it as a short take-home under a clock.
What it tests
- Rapid exploration and modeling on an unfamiliar dataset
- Coding independently without interviewer hints
- Summarizing results and limits clearly
How to prepare
- Practise a 90-minute end-to-end notebook: explore, baseline, improve, report.
- Read up on basic recommender approaches before the final stage.
Long solo data exercise in 3 final-stage reports (2015-2021); not confirmed in recent reports.
08Hiring manager and wrap-up
About 30 min with a manager, director or team; sometimes a separate team discussion
The loop usually ends with a short conversation with a senior team member about your experience, the team and your questions, and occasionally a behavioral round. One candidate first met their own team only at this point. The company's hiring page also lists a hiring manager interview and, in some cases, a final talk with a senior leader.
What it tests
- Motivation and fit with the team's mission
- Collaboration and handling disagreement
- Lessons from a past project, including what you would change
How to prepare
- Prepare STAR-style stories on collaboration and problem solving; the company recommends this.
- Have questions ready about the team, since you may meet it only here.
Closing behavioral or manager conversation in 4 data science reports (one from 2024); the company's hiring page lists hiring manager and senior leader conversations.
Questions and themes reported most often
- Bayes rule with coins or mixed groups (5 reports, math, statistics and probability)
- Hypothesis tests and A/B testing basics (5 reports, math, statistics and probability)
- Predict salaries from a job dataset (take-home) (4 reports, take-home or online test)
- Red and blue balls, with and without replacement (2 reports, technical screen)
- Gradient descent or linear regression from scratch (2 reports, technical screen)
- Merge two sorted lists, then variations (2 reports, coding round)
- Design a job recommendation system (2 reports, machine learning and ml design)
- Handling missing data and overfitting (2 reports, machine learning and ml design)
What Indeed asks about
- Probability & statistics28%
- Coding24%
- ML concepts & system design24%
- Take-home & data exercise16%
- Behavioral & fit8%
Indeed interview prep checklist
- Built a regression model on a mixed-type dataset and wrote up choices and errors
- Solved with and without replacement probability questions without notes
- Explained when to use a t-test, z-test or a non-parametric test
- Wrote linear regression with gradient descent from scratch
- Sketched an end-to-end job recommendation or job-spam design with online testing and monitoring
- Completed a 90-minute solo data exercise from raw data to summary
- Prepared two STAR stories and questions for the hiring manager
Tips
- Do the take-home properly: a clean baseline, honest validation and clear written answers carry more weight than a fancy model.
- Drill textbook probability (draws with and without replacement, Bayes rule) until set-ups are automatic.
- Review hypothesis testing in depth, including t-test vs z-test, p-values and non-parametric alternatives.
- Practise easy to medium coding problems with spoken complexity analysis and follow-up variations.
- Be able to write gradient descent for linear regression from scratch in numpy.
- Prepare ML designs grounded in a jobs marketplace: recommendations, job quality and spam, apply-rate experiments.
- Plan for a long final day of single-topic rounds and pace your energy.
- Expect slow or patchy communication at times; follow up with your recruiter politely.
See the rest of the Indeed loop
The later stages, the questions reported for them and how to prepare are in Pro.
Unlock with ProPractice the Indeed loop
Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.
Indeed interview FAQ
How many interview rounds does Indeed have for ML roles?
Candidates report 8 stages: Recruiter call, Take-home or online test, Technical screen, Coding round, Math, statistics and probability, Machine learning and ML design, Data exercise and Hiring manager and wrap-up. A recruiter screen, then usually a take-home modeling assignment (often predicting salaries from a job dataset) or, in some offices, a timed online test mixing probability, coding and a small ML task. A one-hour technical screen with two data scientists follows, covering easy coding, probability, statistics and ML basics. The final stage is a set of about four focused rounds, each typically with two interviewers: coding, math and statistics, machine learning, and a 60 to 90 minute solo data exercise, closing with a short conversation with a manager or senior leader. The company says interviews are now virtual and that multi-round decisions can take several weeks; candidates report anything from three weeks to four months. Machine learning engineer reports from 2026 describe an ML design phone screen and ML coding from scratch, with a harder data-structure coding round later. This outline rests on about two dozen reports, most of them from 2021-2022 and earlier.
How long does the Indeed ML interview process take?
Reported timelines run 3 weeks to 4 months.
What does Indeed ask in ML interviews?
By reported question mix: Probability & statistics 28%; Coding 24%; ML concepts & system design 24%; Take-home & data exercise 16%; Behavioral & fit 8%. The most repeated themes: Bayes rule with coins or mixed groups; Hypothesis tests and A/B testing basics; Predict salaries from a job dataset (take-home).
How should I prepare for the Indeed ML interview?
Do the take-home properly: a clean baseline, honest validation and clear written answers carry more weight than a fancy model. Drill textbook probability (draws with and without replacement, Bayes rule) until set-ups are automatic. Review hypothesis testing in depth, including t-test vs z-test, p-values and non-parametric alternatives.
Are there free Indeed interview questions?
Yes. LeetML has 9 questions reported from Indeed interviews, and 2 of them are free to solve in the browser; they are listed by stage above.
Related interview guides
- LinkedInProbability-flavoured coding, ML fundamentals and recommender design built around LinkedIn's own products.
- ZillowReal-estate flavoured ML and data loops: a hiring-manager or coding screen, then a mixed loop of coding, ML depth and cases.
- GleanGraph-search coding, an AI-assisted build round and enterprise ML design, judged on a small set of reports.
- WayfairA timed online assessment, then several business-flavoured ML case studies, one coding round and a behavioral round.
- Expedia GroupRecruiter 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.
- Booking.comCase-driven ML and data science interviews on real travel-marketplace problems, with a coding track for ML engineers.