It's tempting to go down the rabbit hole.
There's always another library to learn, whether it's Scikit-learn, Pytorch, Tensorflow, Huggingface, cv2, NLTK, spaCy. There are more popping up, too, so that list is less than incomplete.
When you have a good grasp of some modelling libraries, you should start pivoting.
Get started with ML Ops
Most students learn machine learning in Jupyter Notebooks.
Don't get me wrong. I love Jupyter for what it is. But whether it's notebooks or a collection of scripts, these are difficult to work with in teams and companies. Start learning to save and retrieve models after training.
Then you can go ahead and learn this amazing tool.
Learn Containerisation with Docker
Virtual machines and containers revolutionised computing.
Docker is a tool to back an operating system and your code into a neat frozen state that can be used and reused. You can make this Docker available anywhere, and your colleagues can run your code regardless of their OS.
Understanding the basics of containers will significantly increase your knowledge of ML.
Create Paths into your Docker container
Your Docker is closed by default, but you can change that!
You can even build paths into the container! These endpoints are a great way to think about how you want people to interact with your model. Then you can have other programs interact with your fancy model inside of your Docker container.
This additional knowledge is one of the high-demand skills currently.