is this ML workload an "I trained a model, pickled it, and now we ship that around in production" type situation?