You can design galvanic isolation by just making two circuits that aren't connected to each other.
The difficulty is transferring signals between them without breaking the isolation. The TI parts described in the document you linked are one way of doing that. Optoisolators are another. Transformers are another (for purely AC signals).
How can I start designing this and which simulation programs do I need?
Generally simulators can be useful for predicting how a signaling system works at transferring signals, but they won't do much to tell you whether you've maintained or broken the isolation in your system.
Can the isolator design be realized on PCB or only CMOS?
If you buy the parts from TI (or similar parts from Analog Devices or other vendors) then your signaling across the isolation barrier is implemented in CMOS. But it's also implemented at the board level because you bought the part and put it in your PCB.