I'm not "pivoting to a ML engineer" but in the last 2.5 months I've learned to some extent to use public models, use the tools and APIs to train and run them. That was a lot of reading with little code writing. I didn't pivot into it, that was part of the project (object recognition in a video stream). It helps if you work with small organizations that don't box you into a role but just give you stuff to do.
Could you please share some of the learning resources you used and found useful. I am overwhelmed by the amount of ML related material out there and am having a hard time finding out material what is worth my time as a software engineer.
The resources are all crap to be honest. Half of them have been obsolete for a year or more and most of the rest look like they're done for self promotion and assume you already know everything. Every public repo you run into has already been forked 3 to 10 times and now you have to find out which one(s) is/are up to date.
I had one of these toys:
https://shop.luxonis.com/collections/oak-cameras-1
They're cameras with a colour cam, two b&w cams that give you depth info (z-distance) and a small coprocessor from Intel that can run a reduced neural network directly on the cam.
The other thing they have is their own API and pretty good documentation for it.
It won't teach you about ML math but you'll get used to loading pre trained models and getting your data out of them. And then you'll move on to training your own model (look at DarkMark for example), converting models between various formats and other stuff like that that I'm still learning.
And you get a pretty fun toy!
Mind, i've only worked with object recognition. I have no idea about LLMs or other applications of neural networks.