I am super happy with Erlang/Elixir, having build and run game backends, Kafka replicators, an ML platform (models in Python) and several social apps over the years.
My story is a bit different but maybe interesting to contrast:
I used to work with Ruby for years when I switched to Erlang 9 years ago. So for me the syntax was unfamiliar and the actor model was completely new. Elixir and Phoenix where not around so I started with processes, messages, gen_server and the like. I started building an Erlang-process load balancer, learning about mailboxes and supervision in the process. Later I switched to Elixir and felt reminded of the Ruby syntax, really liking it plus new things like the pipe operator.
For me the biggest learning that needs to happen (apart from general FP) is on structuring applications. One way to think about it is having long-running micro services within the VM along with short-lived processes for small tasks like serving a single http request.
I can see that coming from Ruby today and using Phoenix can give you a cozy feeling where you build stuff without thinking too much about OTP, processes and how the VM does things (scheduling, mailboxes, monitoring, GC, supervision, distribution). And I think that's totally fine unless you want to step outside the framework and really make use of all the things that set Erlang/Elixir apart.
So maybe one way to get started today is looking at Erlang first, explicitly because it's unfamiliar.