Generally there are 2 ways to design for this problem:
Once you have an actor reference, you can send things to it. So you can simply do this:
targetActorRef ! myThing
When the actor is done with the processing, it will call:
myThing.myMethod
This design is OK when the target actor is very reliable and it lies in the same node with myThing.
Another way, is to use an actor to wrap around myThing. This wrapper actor will communicate with targetActorRef. It may watch targetActorRef to handle the case when targetActorRef dies, targetActorRef doesn't respond after some timeout etc.