(FEAT) Ability to find ideal inference engine

I know it's big feature request, but if it can be done it will benefit community immensely.
Basically figuring out automatically which inference engine+settings+quantization for any model is going to be ideal for system.
Will need going inference engines one by one, which will be lot of work. But thing is, lot of people do it manually countless times. If it can be automated, it will save countless individual hours.