student trying to break into interpretability research, where do I even begin

fascinated by the idea of actually understanding these models. the field looks intimidating from outside. would love a realistic path in from people already doing it.