I'm sure this is due to my beginner status in ML/DL, but I'm really disappointed in how much deep learning seems removed from the things that I enjoyed the most

I'm sure this is due to my beginner status in ML/DL, but I'm really disappointed in how much deep learning seems removed from the things that I enjoyed the most in statistical learning. I enjoy the creative challenge of applying domain knowledge when building (for example) linear or bayesian regressions. In contrast, DL seems like a whole bunch of hyperparameter tuning and curve plotting. Curious to see if this assessment seems correct from those more experienced...