You might like having a go at Lush. It has fallen out of favor of late but is a very interesting language/system.
https://scottlocklin.wordpress.com/2024/11/19/lush-my-favori...
Sounds interesting but I'm using very spare very high rank tensors, e.g. rank 3 neuron equivalents.
As such pretty much all numerical optimisations are useless for my work. Racket however chugs along happily, if slowly.