Find some detailed job ads, and work backwards from there to see what skills are in demand.

What exactly do you mean by "infrastructure" or "systems"? At my organization, that basically means stuff like grafana, open telemetry, kubernetes, aws/azure, etc.

Your "AI infrastructure" is perhaps what other people call "MLOps" - data pipelines, airflow, that kind of stuff.

Both of those are different from each other, and also different from being a developer/SWE etc.

Its a bit annoying to me how these things have become segmented, I kind of preferred when it was all just "programming" or "computers" and most people did most things, but unfortunately these days the market is quite segmented / specialized, and the key to getting any particular job is to already be familiar with the popular tools in that space.

I'm curious, what's the reason for you to move toward more system-oriented work? Is it because you're looking for a way of career progression?

If so, I'd suggest you have some in-depth conversation with your manager and your mentor (if you have one), to identify area of skill developments for you, and opportunities for "learning on the job". And if your current situation/management won't afford you that opportunity, you should focus on finding a new job, which should be fitting for your current skill set, but with clear space for doing something more.

And this is what I am doing right now, like I am just reading the DDIA book and writing the raft consensus algorithm in Go just for learning purposes