You might like these posts on gradient free optimization:

http://pchiusano.github.io/2019-07-06/learning-without-a-gra...

http://pchiusano.github.io/2020-10-12/learning-without-a-gra...

I think you’re right that the area is underexplored.