Handling of intents containing unique first or last names

You can use a STT engine that does custom voice models. There’s two major ones I know of you can check out:

Kaldi
Deepspeech

Both will take a good deal of tweaking and testing, I presume, to meet your needs.