Experimenting with a workflow of training a model with Python in a Jupyter notebook and making a Rust script to serve it up. Quite fast already, even with no optimizations!
real 0m0.280s
user 0m0.047s
sys 0m0.219s```
```time python make_predictions.py
real 0m2.388s
user 0m2.266s
sys 0m1.859s```