Upskilling at the job - how to determine what is relevant?

I (34F) am a software eng at a big tech company. I want to stay up to date with latest stuff happening. One of my sister teams is an ML team and I am helping them productionize some of their stuff, looking into latency problems, etc and I think it's a good opportunity to upskill. I also talked to my manager about upskilling and she agreed she will put me first if any ML work comes up. At my work, I use a lot of Claude, try out new mcp tools, play around with workflows, watch YouTube videos of people sharing their workflows, etc but it all feels really random.

I used Claude to create an AI upskilling plan for myself without going into the hard core math of the LLM. The plan is divided into 8 steps (overview of LLMs, prompting & context engineering, and 6 more) where each step is couple of weeks. I think couple of weeks for each step is a little too much. Before going through all of the material I double check if this material is really useful by poking Claude again. So far I'm enjoying the curriculum and things are making more and more sense - just finished watching 3.5 hours of Andrei Karpathy LLM overview.

But I guess I'm being a little indecisive right now, is it worth spending all this time in an organized way, should I take some more shortcuts, is it even useful studying all this, etc. for example, would it even make sense to do a 2 hour exercise on building a small chatgpt? I want to hear from other experienced devs what has their learning been like and are you setting any goals around learning? What are you prioritizing?