Hi, I have 10+ years of experience as a SWE, including some projects where I did some ML work: recommendations of some kinds to users, figuring out the best features to use, what data to bring in the training set etc. I did this pretty naturally/intuitively and only realized later on in a book that this was what I was supposed to do.

I also come from a strong analytical/math background

I was curious if anyone went the SWE -> ML route? and what was their experience / if they regret are not etc?

I managed to get an offer at a FAANG and I have the impression that ML meant tuning model parameters to have a 0.1% increase of anything (when things are mature), and doing normal SWE work (pipelines to bring features to a model) when things are less mature

On the plus side I have the impression that there is way less oncall work in ML

On the downside I don't know if people get satisfaction from the job? Also I was curious if the experience was significantly different in a big company (e.g FAANG) or in a smaller company (e.g 1000 people)

Thanks in advance!