Software Engineer - Machine Learning applicants have rated the interview process at Google with 3.4 out of 5 (where 5 is the highest level of difficulty) and assessed their interview experience as 40% positive. To compare, the company-average is 61.5% positive. This is according to Glassdoor user ratings.
Candidates applying for Software Engineer - Machine Learning roles take an average of 49 days to get hired, when considering 10 user submitted interviews for this role. To compare, the hiring process at Google overall takes an average of 38 days.
Common stages of the interview process at Google as a Software Engineer - Machine Learning according to 10 Glassdoor interviews include:
Phone interview: 56%
One on one interview: 33%
Skills test: 11%
Here are the most commonly searched roles for interview reports -
I applied through an employee referral. The process took 6 weeks. I interviewed at Google in Oct 2018
Interview
Classical leetcode problems mixed with some hardcore OS multithreading questions. Everything on the whiteboard. Nothing out of the ordinary, but very strong grasp of multithreading is expected. Some of the interviewers were not excited about their job, which did not make a good impression
Interview questions [1]
Question 1
Classical computer science questions, cannot talk about specific, NDA
I applied through a recruiter. I interviewed at Google (New York, NY)
Interview
4 rounds (online) - 2 coding, 1 ML systems, 1 behavioral. Coding were standard leetcode style. ML system was expected. Behavioral was also the standard questions. All went pretty smoothly
Screening interview is about a coding problem. It was a 45 mins interview. I had no question about machine Learning. It was only about coding and software engineering. The mistake I did is that I spent too much time explaining my approach on the first question. While the interviewer had more questions for me.
Classical google interview process;
HR phone call,
Algorithm interview and other technical interviews.
Good to have a feedback after every step from hr. However unnecessarily difficult algorithm problems are annoying, after 10 years they are still doing same, no improvement.