[apohn]

>I did not get the position, but I would recommend you look elsewhere if your focus is truly machine learning

For ML, it's really important to join a team where your interests in ML are aligned with what the team does. It's really hard to see this from a job description. There's enough going on at Google that you can find work that fits.

I've interviewed twice at Google and had the same experience as you. No ML or math questions at all. More algorithms and how to quantify a business problem. That being said, I asked enough questions to realize that some groups use ML, but that's a small part of what they are doing. For example, they might have a platform for doing A/B testing and the "Data Scientist" job is really defining A and B and feeding that into the platform to extract metrics. How much ML being done is going to be different on the Ads team than a customer facing services role for Google Cloud.

I had similar experiences interviewing at Facebook, just with more probability and stats brain teasers. Facebook doesn't guarantee which job you'll get once you're in. You go to the bootcamp and then which team you end up is decided after the bootcamp. That doesn't work for everybody if there are certain types of ML work you're not interested in.

[isuckatcoding]

Wow that sounds like such a boring job.

[apohn]

For the the type of role I described in my post, being able effectively define metrics to quantify a business problem (e.g. how can we better engage with group X on our platform) and work with product engineering to build those features (including helping to define the data collected and how it's stored) into the platform is more important than tweaking a bunch of parameters in a machine learning model.

A lot of those folks are not only thinking of ways to quantify a business problem, they are actually thinking of new business problems (e.g. what does it even mean to "engage" on our platform). It can be quite creative and challenging.

Unless you are writing the core algorithms or working as a statistician, a lot of ML jobs are some variation of the above - basically coming up with ways to turn business problems into data/features you can feed into a model and picking an appropriate model. How much code you write and the tools/languages you are using will depend on the job and size of the company.