Amazon switch from SDE to Applied Scientist
I am wondering what is the process of switching from SDE 1 to an Applied Scientist at Amazon. I got an offer for SDE 1, but I'm more interested in Machine Learning. I have a master degree with an emphasis on machine learning. Also how difficult is it to make the transition?
Don’t get fascinated with the “scientist” word in the title. They may well be folks who can’t code but try to crank out useless models for the sake of fake credits. I have a CS PhD and have done postdoctoral research in a highly respected research lab before working as a software engineer many years ago. I love coding and I develop data model too, but I don’t need a scientist title to boost my confidence
I would just want to develop models and continue to read papers about machine learning. As an SDE I have a high chance of landing a team that does mostly front end.
I have worked in mostly science heavy groups for the last two years (I'm an SDE3 but have a CS PhD and had an academic career before Amazon). All of the ASes I've dealt with (> 30) have PhDs and have strong expertise and a good publication record. I don't know if all AS roles require a PhD but likely very difficult without one. I think many if not most SDEs have an MS and everyone loves ML-- so while it might be possible, you will face strong competition. It might be better to go for SDE roles in ML focused teams. I know many SDE2s that gunned for our team for that reason. Internal transfers are easy but ML teams are in high demand. Our team doesn't take any SDE1s and we've had dozens of SDE2 internal candidates that we've passed on that sound just like you MS, ML enthusiast, does kaggle, etc. I think of amazon as just a microcosm of the real world but the skills distribution is just shifted up a bit-- there are clusters of super talented people and super desirable teams then most are ok and plenty are bad (teams and people, you hear them complaining on Blind the most). So it's great to get into Amazon- congrats! But now you have to compete to get into the best teams. It never ends :)
Why do SWEs want to implement ML models that ASes build? It’s not really fun is it?
I think it varies. For me it hasn't been such a clean break like: AS does all interesting work -> SDE does grunt integration. In my experience (and many others I've known and worked with) the ML engineers are involved very early, contribute to ideas and model building, and there are many interesting algorithmic problems that pop up when trying to do interesting ML work at scale that the ASes often aren't solving. It's usually been quite collaborative. I still publish papers and present at conferences too so it's not just boring typing. That being said I'm sure there are plenty of ML engineering projects (like many engineering projects overall) that are boring. *shrug*