may be micro contact printing. spinning coating PDMS on a glass, and use a PDMS stamp smaller than well plate to remove the PDMS layer on the surface of PDMS. but the thickness could not be guaranteed.

Maybe You could spin-coat each coverslip with the right PDMS mixing ratio and cure it before placing them inside the wells plate. By the way, it is probable that the PDMS with very low curing agent content will not reticulate enough to prevent further spreading inside the well. I believe "50 microns" is a way to say "thin, controlled layer"?

Unfortunately achieving absolutely uniform 50 um coating within each well would be quite tricky. Mostly because of meniscus effects, where PDMS in the center of the well will be thinner than near the walls (similar to what you might see when collagen solidifies within the wells). If you can live with meniscus and look only at the center, then gravimetric approach would be the easiest (simply compute the mass of the PDMS volume that on average would give you 50 um in the well).

A more complicated approach would be to start with the bottomless well-plate. Spin-coat your 100:1 PDMS ratio onto a glass-slide. Then place your bottomless well-plate on top (might need to use a clamping arrangement) and them add required volume of the curing agent to other wells with the precision pipetters. You'll have to figure out the appropriate increments and curing protocols (maybe do it at RT for 24 h and then 2 h in 65C oven)

Another way is to spin each ratio separately, then cure, punch out the disks, place them onto the glass slide, and then place bottomless well plate on top. Now, the practical limit of what you can handle with tweezers is between 15:1 and 20:1 ratios. Anything beyond that is way too gooey.