[gevorg_s]

Aim easily handles 100s of ML training runs. There are folks who have 1000s of experiments too. We haven't oficially benchmarked it yet though.