The right way to do this is to use the API node, which is a lightweight dedicated node for scoring a single observation (or a couple of them). You create an API service in DSS with your python model, package it & import it into the API node, and you can score with a JSON API.
Check out the documentation: Real-time predictions.
On the other hand, the DSS public API (of the design node & automation node) lets you control DSS itself: export data, change configuration, launch jobs, etc. but is not meant to be called for individual scoring.