Definitely. Everyone is still painfully assembling systems based on open source libraries many of which have their origins in research. It‘s not enough to assemble „ML platforms“ that integrate with as many of them as possible.