This repository is a personal project I made to deploy an MLOps pipeline on Cloud Composer (Apache Airflow):
-
The /dags folder contains the Airflow dag that trains a random-forest algorithm every 2 weeks, when train data is shared on Cloud Storage (via
GoogleCloudStorageObjectSensor). The script then deploys the newly trained model on AI Platform. -
The /function-source folder contains the source code to deploy the ML model over a Cloud Functions endpoint. The API is secured via an API key.