Yelp Product Data Scientist Job Interview

Historical candidate account: this article preserves one person’s reported experience. The interview date was not recorded in the original text, and updating this page does not make the account a current hiring process. Names, levels, durations, questions, outcomes, and approximate compensation are as reported, not independently verified or endorsed by the employer. Confirm the current role and assessment with the recruiter.

Jonny, an instamentor mentee, reported completing a virtual onsite interview at Yelp and received a great job offer as a senior data scientist.

Here is what Yelp's virtual onsite interview process looked like.

Yelp Data Scientist Job Interview Summary

Candidate: Jonny

How it gets started: applied on LinkedIn

Job Level: Senior

Year of Experience: 5 years experience 

Degree: M.S in computer science

Offer: Yes

TC: ~200K USD

Location: San Francisco, CA

Interview process: 3 weeks

Preparation: 1 month

Has a job: yes

Decide to join: N/A

Yelp Data Scientist Job Interview Round 1: recruiter chat

Basic questions such as salary expectations, whether willing to work in the SF in the city in the future. What do you know about Yelp?

Yelp Data Scientist Job Interview Round 1: Hiring Manager informational chat

Introduce each other, what the role is about, which other teams you will be collaborating with, and example ML projects that you will be working on. 

Yelp Data Scientist Virtual Onsite Interview

Round 1: behavioral questions 

1. Why Yelp?
2. How do you handle conflict with your coworker?
3. Give me an example of a challenging project that you recently delivered.

Round 2: Python and Product Sense

1. Basic pandas data processing question with a test data set;
2. How can we get more businesses interested in listing at Yelp?
3. How long do we need to run this A/B testing on a new restaurants recommendation algorithm on Yelp? 

Round 3. machine learning and product sense

1. Difference between logistic regression and neural network.
2. Why boosted tree usually work better than a simple logistic regression model?
3. How do you choose a model with similar performance?
4. Feature selection and regularization.

Round 4: SQL and product sense

1. How do we evaluate a new messaging feature on Yelp for business and customer communications?
2. Write a SQL query to compute the metrics in #1.
3. How do we test two different versions of user interface designs? How can we improve the test?

Offer

Jonny received the job offer after two weeks of the virtual interview. 

Connect a model comparison to a product decision

Boosted trees do not universally outperform logistic regression. Compare the models on the same appropriate data split, examine uncertainty and calibration, and consider latency, interpretability, maintenance, and failure costs. For a messaging feature, define successful business-customer exchanges, response time, spam or abuse, and eligibility. Use the experiment’s assignment unit and full observation window when estimating effects. A more complex model or a higher engagement count is not sufficient evidence of a better user outcome.

Practice one reported prompt, state the data and assumptions, produce a checkable answer, and explain one limitation. Then change a condition and see whether your reasoning still holds. Use this related practice resource for a follow-up; it is separate from the historical interview report.