Hi Akrys,
It’s hard to provide a concrete solution without knowing what kind of complex we are talking about. 
But for facial rigs it’s usually boils down to controlling said vector1s (anim curves). Whether it’s blendshapes, or material curves you can (as you know) drive them from a ControlRig.
Metahumans do it in similar way, they take control positions and remap them to expression curves, before feeding all curves into RigLogic which outputs correct bone transforms, blendshapes and material curves for wrinkle maps.
You most likely can breakdown your rig into curves as well, whether you following FACS system or ARKit, or with phonemes etc.