Then I use a special tool (really just a mouth tube that has a glass tube at the end) to single tardigrade to a glass-bottom petri dish that goes into the scope. I take movies, then label ~hundreds of frames manually with "tardigrade" "algae" and "junk" then fine-tune an existing object detector with those frames. The centroid of the object detection provides the direction to move the stages to center the detected object.
The other mode is fast scanning, which I described above (move the camera continuously over a large field of view, pulsing the LED quickly/periodically and dropping any empty frames).
(at work I get to play with grown-up microscopes and they are another class entirely; precision machines with extraordinary capabilities and price tags to match)