DataRobot supports 2 "portable prediction" options for using your model outside the platform:
Scoring Code export which is a Java JAR implementing the same model (not an approximation) which can be executed with a simplified Java- based runtime.
Portable Prediction Serverwhich provides a docker-based runtime image and you can export the leaderboard model as a DataRobot Model Package (mlpkg) to run within the docker container and serve predictions.
Several of these options require an MLOps license, so make sure to take note of that in the documentation and reach out to your account manager or support team if you need to upgrade your license to take advantage of this functionality.
DataRobot Prime model provides you with the code, it is done for the purpose that pickle produces different files under different environments. So we can't ensure that generated pickled file will be supported by your software and hardware. Meanwhile, by providing you with a ".py" file you are able to pickle it yourself, so you will be sure that it works with your setup.