I just learned that Pytorch is now supporting java for inference. So we could use lib python-CLJ for training and directly call the model from Clojure for production.

Do you mean JPMML?

No really pytorch 1.4

as inspired at the meetup in the mob session in Santiago’s group: using clojure spec to validate input data and generators to clean it up: https://nextjournal.com/kommen/clojure-spec-for-data-science

Thanks for the great Clojure DS meetup today! Was a pleasure to meet you and have a chat!