We are building an application that uses machine learning to predict certain parameters and use them in an optimization algorithm afterwards.
We are not building the front-end in R neither the back-end and our only R component is the feature engineering, training, serialization and prediction scripts that generate a couple of .RDS files and expose them via REST-API using AWS sagemaker endpoints.
- The feature engineering is done with dplyr (MIT license), lubridate (GPL2/3) and forecastML (MIT license).
- The training is done using keras with tensorflow backend (Apache license), glmnet (GPL2/3) and xgboost (Apache License).
- Prediction is done with the same libraries and with MLflow (MIT license) which exposes this model to sagemaker
- Serialization is done of course with base R.
If my end-user is an external company that is going to take the results of the optimization (which are consequently feed with the machine learning predictions) do I have a problem in terms of licensing?
At the end the end-user does not receive any code it only receives the predictions via an API.