Ask HN: Transitioning to ML Systems and Performance Work: Advice Needed

I hope you're all doing great! I'm an ML Engineer currently working at a FAANG company, where I primarily focus on data science analysis and basic model building. However, I'm eager to switch gears and dive into the exciting world of ML systems and performance work. I want to work on optimizing the DNN stack, developing training/serving/inference systems, and delving into low-level ML compilers. Here's the catch: my experience in this area is limited, as I last worked with simple DNNs during my undergraduate and master's studies about five years ago.

I'd love to hear from those of you who have successfully made a similar transition or have valuable insights to share. How can I effectively pivot my career towards ML systems and performance work? What steps should I take to acquire the necessary skills and knowledge? Are there any specific resources, courses, or projects you would recommend for gaining hands-on experience in this field?