Company interview guide
Yelp Machine Learning Interview Guide (2026)
Practical, mostly easy-to-medium coding with a text-processing flavor, an ML design round and two values interviews.
Yelp's machine learning interview loop has 7 reported stages, from online coding assessment to behavioral, past projects & hiring manager; typical length: 4-8 weeks.
Roles covered: ML Engineer and Data Scientist. The question bank holds 3 questions reported from Yelp interviews, 1 of them free to solve in your browser.
- Interview process length
- 4-8 weeks
- Stages
- 7
- Reports linked to bank questions
- 5
- Bank report dates
- Jul 2017–Apr 2025
- Questions in bank
- 3 (1 free)
- Last reviewed
- October 2026
How the Yelp interview loop works
Most ML-engineer candidates start with a timed online coding test that arrives soon after they apply, sometimes before any human contact, then a 30-minute recruiter call (a few report the call first). Next comes one 45-60 minute live coding screen with an engineer. The final loop is about four 45-minute interviews in one sitting of up to four hours: one coding, one ML system design, and two behavioral or past-project conversations, one of them often with a manager. Data-science candidates get an online test or recruiter call, a technical screen on SQL, Python, statistics and a product or modeling case, and a final round with data scientists, product managers and the hiring manager. The company's engineering blog says its interviews are structured and scored with a points rubric, and that technical questions are scaled-down versions of problems its teams have solved. Only about eight ML-engineer reports and one data-science report date from 2024 onward, so treat the outline as a guide.
The Yelp interview process, stage by stage
01Online coding assessment
Automated HackerRank test, about 30-60 min, one or two problems; data-science versions include SQL
Often the first contact after you apply, sent without a call. Difficulty varies widely: some candidates got an easy 45-minute problem, while two 2024 ML-engineer candidates got a hard problem (one a bit-manipulation task) and failed on hidden time limits. One 2024 test paired a SQL query with a problem best solved with a heap.
What it tests
- Correct, efficient code that passes hidden tests within time limits
- Everyday data structures: dictionaries, heaps, sorting
- SQL joins and aggregation (data-science and some ML versions)
- Python data manipulation
How to prepare
- Check the complexity of your solution before submitting; timeouts on hidden cases are the common failure.
- Practise in the HackerRank editor so the input format and test runner are familiar.
- Expect anything from easy to hard and budget time for an optimisation pass.
An online coding test is the first technical step in 7 ML-engineer, 5 data-science and 1 software-engineer report; the company's engineering blog names a code test as the first step of its process.
02Recruiter call
About 30 min with a recruiter, before or after the online test
A conversation about your background, the role and logistics. Candidates describe recruiters who lay out the upcoming rounds clearly; older data-science reports include a few statistics or SQL questions in this call. One 2026 report also had an optional 15-minute culture chat.
What it tests
- Why Yelp and why this team
- A short summary of your relevant work
- Timeline, location and level
How to prepare
- Ask for the type and length of every later round; the invitation emails are usually detailed.
- Ask whether the final loop runs in one sitting, since several candidates found four hours back to back tiring.
- Data-science candidates should be ready for a light stats or SQL question here.
A recruiter or HR call appears in 4 ML-engineer, 6 data-science and 1 software-engineer report, mostly about 30 minutes; in ML-engineer reports it usually follows the online test.
03Live coding (ML-engineer track)
One 45-60 min screen with an engineer, plus one more coding round in the final loop; shared editor such as HackerRank or CoderPad
Problems are practical and usually easy to medium: text processing with specific rules, counting and ranking items in lists, building pairs from a set under conditions, or a weighted random sample. Tokenizing text and splitting it into bigrams has come up in at least three loops. Interviewers care about how you clarify requirements before coding as much as about the final answer.
What it tests
- Turning loosely stated rules into correct string and list code
- Dictionaries, sorting and simple ranking
- Tokenization and n-gram handling for NLP-flavored tasks
- Random sampling with weights
- Talking through edge cases and complexity
How to prepare
- Ask clarifying questions before you write code; rejected candidates were told they asked too few.
- Practise tokenizers with explicit rules (punctuation, case, contractions) and bigram counts in plain Python.
- Write small helper functions and test them on an example you choose yourself.
Free reported questions for this stage
- Byte-Pair Encoding Tokenizer from ScratchML CodingHardSolve
Live coding in 10 ML-engineer reports, 5 since 2024, consistently easy to medium; the company's blog says first-round technical questions are scored on how many parts of a problem you complete.
04Technical screen (data-science track)
One or two 45-60 min video rounds with a data scientist: SQL, Python or pandas, statistics and a product or modeling case
Data-science screens mix several SQL questions and a Python task with statistics and a case built on a Yelp feature, such as predicting waitlist times or review likes, or defining success metrics for an event. Interviewers dig into statistics and into the ML models on your resume. Most of these reports predate 2023.
What it tests
- SQL joins and group-bys of medium difficulty
- Hypothesis testing, A/B test setup and metric definition
- Framing a prediction problem: features, model, evaluation, business value
- Explaining the math behind models you have used, such as boosted trees
How to prepare
- Prepare to define a primary metric and guardrails for any Yelp feature you can think of.
- For modeling cases, cover features, a baseline, how you would validate it and how you would value the gain.
- Know your resume models mathematically, not just by library name.
A technical screen with SQL, Python, statistics or a case appears in 13 data-science reports, but only 5 date from 2022 onward and just 1 from 2024.
05ML system design
About 45 min in the final loop, open discussion
The named prompts are recommendation and ranking systems. Two candidates found this the least collaborative round, and one was told afterwards that they had not asked enough questions here. Older data-science screens also included open-ended modeling or design questions.
What it tests
- Scoping a vague product goal into a concrete ML problem
- Recommendation and ranking design: candidates, features, ranking, feedback
- Choosing offline and online metrics
- Trade-offs between simple and complex models
How to prepare
- Spend the first minutes on questions about users, data and success criteria before proposing anything.
- Have a recommendation-system blueprint ready, adapted to local businesses and reviews.
An ML design round is in the final loop of 6 ML-engineer reports and 1 data-science report; the company's engineering blog lists system design as one of its standard interview types.
06Final round (data-science track)
Several interviews of about 45 min with data scientists, product managers and the hiring manager
Few reports describe it. They mention product and metric questions, A/B testing, statistics and ML, plus behavioral conversations, with meetings across data science and product.
What it tests
- Product sense and metric design for Yelp features
- Experiment design and interpretation
- Statistics and ML fundamentals
- Working with product managers
How to prepare
- Prepare to explain an A/B test result to a product manager in plain words.
- Ask the recruiter for the round list; the reports here are too few to pin it down.
Only 3 data-science reports describe the final round, one from 2024; they agree on product, experimentation and statistics content but not on format.
07Behavioral, past projects & hiring manager
Two 45-min conversations in the final loop, often one with a manager; some loops have a separate past-projects round
The company's engineering blog describes two standard values interviews, one on ownership, tenacity and curiosity and one on working well with others. Candidates report questions on failure, conflict, teamwork and why Yelp, and a deep dive on the project you are proudest of.
What it tests
- Ownership and persistence on a hard problem
- Collaboration and handling disagreement
- Depth behind a past project you led
- A genuine reason for choosing Yelp
How to prepare
- Prepare stories for failure, conflict and an uncomfortable situation at work; these recur.
- Pick one project you can explain end to end, including what you would change.
- Map each story to ownership or to working with others, the two values the company scores.
Behavioral, project or hiring-manager rounds in 7 ML-engineer, 3 data-science and 2 software-engineer reports; the company's blog names two values-based interview types.
Questions and themes reported most often
- Why Yelp, and a deep dive on the project you are proudest of (7 reports, behavioral, past projects & hiring manager)
- SQL plus Python in the data-science technical screen (7 reports, technical screen (data-science track))
- Easy list and dictionary tasks: count items, rank, pair or filter records (6 reports, live coding (ml-engineer track))
- A/B testing and defining success metrics for a Yelp feature (5 reports, technical screen (data-science track))
- Tokenize text under specific rules and split it into bigrams (3 reports, live coding (ml-engineer track))
- Design a recommendation and ranking system (3 reports, ml system design)
- Failure, conflict and an uncomfortable situation at work (2 reports, behavioral, past projects & hiring manager)
- Hard timed online problem where hidden tests time out (2 reports, online coding assessment)
What Yelp asks about
- Coding (strings, lists, dictionaries)35%
- ML system design20%
- SQL, statistics & product case20%
- Behavioral & past projects15%
- ML concepts10%
Yelp interview prep checklist
- Solved a timed online problem with attention to worst-case runtime
- Wrote a rule-based tokenizer and bigram counter in plain Python without hints
- Practised asking three or four clarifying questions before coding
- Can sketch a recommendation and ranking system with offline and online metrics
- Have stories ready for failure, conflict, teamwork and the project you are proudest of
- Have a specific answer to why Yelp
- Data scientists: can define success metrics and an A/B test for a Yelp feature
Tips
- Ask clarifying questions at the start of every technical round; it is the most common feedback rejected candidates received.
- Drill small Python text tasks: rule-based tokenizers, bigrams, word counts and Jaccard similarity.
- Do not underestimate the online test; hidden time limits have failed correct but slow solutions.
- Prepare one recommendation and ranking design adapted to local businesses, reviews and search.
- Expect the final loop in one sitting of about four hours and plan breaks and energy accordingly.
- Frame behavioral stories around ownership and around working well with others, the two values the company interviews for.
- Data scientists: rehearse SQL, a pandas task, a hypothesis test and a metric definition in one hour.
- Feedback after rejection is usually brief, so ask the recruiter what each round scores before you start.
See the rest of the Yelp loop
The later stages, the questions reported for them and how to prepare are in Pro.
Unlock with ProPractice the Yelp loop
Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.
Yelp interview FAQ
How many interview rounds does Yelp have for ML roles?
Candidates report 7 stages: Online coding assessment, Recruiter call, Live coding (ML-engineer track), Technical screen (data-science track), ML system design, Final round (data-science track) and Behavioral, past projects & hiring manager. Most ML-engineer candidates start with a timed online coding test that arrives soon after they apply, sometimes before any human contact, then a 30-minute recruiter call (a few report the call first). Next comes one 45-60 minute live coding screen with an engineer. The final loop is about four 45-minute interviews in one sitting of up to four hours: one coding, one ML system design, and two behavioral or past-project conversations, one of them often with a manager. Data-science candidates get an online test or recruiter call, a technical screen on SQL, Python, statistics and a product or modeling case, and a final round with data scientists, product managers and the hiring manager. The company's engineering blog says its interviews are structured and scored with a points rubric, and that technical questions are scaled-down versions of problems its teams have solved. Only about eight ML-engineer reports and one data-science report date from 2024 onward, so treat the outline as a guide.
How long does the Yelp ML interview process take?
Reported timelines run 4-8 weeks.
What does Yelp ask in ML interviews?
By reported question mix: Coding (strings, lists, dictionaries) 35%; ML system design 20%; SQL, statistics & product case 20%; Behavioral & past projects 15%; ML concepts 10%. The most repeated themes: Why Yelp, and a deep dive on the project you are proudest of; SQL plus Python in the data-science technical screen; Easy list and dictionary tasks: count items, rank, pair or filter records.
How should I prepare for the Yelp ML interview?
Ask clarifying questions at the start of every technical round; it is the most common feedback rejected candidates received. Drill small Python text tasks: rule-based tokenizers, bigrams, word counts and Jaccard similarity. Do not underestimate the online test; hidden time limits have failed correct but slow solutions.
Are there free Yelp interview questions?
Yes. LeetML has 3 questions reported from Yelp interviews, and 1 of them are free to solve in the browser; they are listed by stage above.
Related interview guides
- DoorDashTimed SQL and marketplace case studies for data scientists; project deep dives, coding and infra-heavy ML design for ML engineers.
- InstacartModerate coding, deep questions on classical ML and your own models, and grocery-flavoured ML design and experiment cases.
- AirbnbMarketplace ranking design, a hard look at your past models, one practical coding round and a values check.
- EtsyMarketplace-flavored data science loops built on A/B testing, product metrics and live SQL, plus coding and ML basics for ML engineers.
- PinterestRapid-fire ML fundamentals, edge-case-heavy coding and embedding-centric recommendation design.