Play the long game. There’s plenty of people who don’t start on their career defining works until they are “old”. A particular example is Maziar Raissi. He graduated with a BS I applied math in 2008. As far as I can tell he probably didn’t touch ML until 2016 or so, and from 2017-2020 or so he published several seminal works on Physics Informed Neural Networks. I think he would have been around 30 years old in 2017. Granted he came from an applied math background which definitely was a part of his success, but he didn’t publish anything in ML until later in his career.