No, I haven't, because I don't want to support users with something I'm teaching myself with (it changes quickly as I learn that I did something wrong). In short, I use aluminum extrusion on optical breadboard to build a rigid frame and then mount linear stages based on stepper motors, smooth rods and 3d printed bearings and mounts. This forms a 3-axis stage to move the sample (could be a slide, or a dish, or whatever) which I control using FluidNC (normally intended for CNC) and some hand-written Python scripts for the UI/camera interfacing. It uses industrial vision cameras.
Most components are fairly inexpensive- at least compared to scientific systems- but aren't cheap for a typical hobbyist. The breadboard is >$200, the extrusion is another $100, the cameras at least $250, and my total personal time investment (converted to salary dollars) would be ludicrous.
The important advantage of my design is that I control the closed loop camera/motor control stack (with machine learning inline), and I can raster scan to make large field of view stitches with the camera moving constantly and super-short light exposures.
Here's an example video tracking a tardigrade (keeping its snout as close to the center as possible) I made with the previous generation: https://www.youtube.com/watch?v=LYaMFDjC1DQ