Google Cloud Data Scientist Job Interview

This historical account describes Jay’s reported Google Cloud data scientist interview. The interview date is not recorded here, and the questions and sequence should not be read as Google’s current process.

Google Cloud Data Scientist Job Interview Case Study Summary

Candidate: Jay

Team: Google Cloud

How it gets started: recruiter reached out on LinkedIn.

Job Level: Entry to Mid

Year of Experience: 2 years experience 

Degree: M.S in machine learning

Offer: Yes

Reported historical compensation: approximately US$280,000. Components, vesting assumptions, and time basis were not recorded; the figure is unverified and is not a current salary benchmark.

Location: Mountain View, CA

Interview process: 3 months

Preparation: 3 months

Has a job: yes

Decide to join: N/A

Google Data Scientist Job Interview Round 1: Recruiter Phone Call 

1. Logistics: VISA, salary expectation, start date, ideal team.
2. Team introduction, what kind of project this role will be working on.

Google Data Scientist Job Interview Round 2: Technical Phone Screen

1. How do you decide which google apps (e.g., maps, weather, news) to show by default on google's home page;
2. What techniques do you usually apply to deal with missing data?
3. What methods do you use for dimensionality reduction? 
4. How do you regularize the logistic regression model?

Google Data Scientist Job Interview Virtual Onsite Round 1: Statistics and machine learning

1. How do you explain the confidence interval to non-technical people?
2. How do you determine the duration of A/B testing?
3. What is a null hypothesis?
4. What is statistical power? How do we improve it?

Google Data Scientist Job Interview Virtual Onsite Round 2: Googliness 

The original account labeled this round “Googliness.” The questions below are Jay’s reported recollection; this article does not provide an official Google definition or assessment rubric.

1. Tell me an example that you showed initiative;
2. Give me an example of how you handled conflicts, stood on your ground, and spoke up for yourself.
3. What is your ideal/dream team?

Google Data Scientist Job Interview Virtual Onsite Round 3: Statistics coding

1. Write a function (could be any programming language) to create an unbiased coin, given a biased coin with the probability of tail unequal to the probability of head.
2. Calculate the median of an array without a median function.
3. Time series analysis: what is ARIMA? How do you handle Heteroscedasticity?

Google Data Scientist Job Interview Virtual Onsite Round 4: Leadership/Behavioral questions

1. Why Google?
2. Why do you want to leave your current position?
3. What is your biggest accomplishment/proudest project?
4. Do you have any questions for me?

Offer

Jay received the job offer after one week of the virtual onsite interview.

A worked reasoning prompt: the biased coin

Assume independent tosses with the same fixed head probability p, where 0 < p < 1. Toss twice: map HT to one outcome and TH to the other; discard HH and TT and repeat. The accepted pairs have equal probability p(1−p), so the returned result is unbiased. This argument does not apply unchanged to dependent tosses or changing probabilities. A perfectly one-sided coin never yields an accepted pair; an implementation should define a maximum-attempt failure instead of silently returning a biased default.

For the other prompts, practice stating assumptions before choosing a method. A confidence interval is produced by a procedure with a specified long-run coverage under its assumptions. For time series, distinguish trends and serial dependence from changing variance. These explanations are editorial practice guidance, not claimed answers supplied by Jay.