Short ml examples for data scientists using Google Cloud Platform (GCP).
- Identify fraudulent transactions in a credit transaction database [code]
The approach followed while solving the ML task adopts the following increasing levels of complexity.
Low complexity; service use
- Training and prediction (BQML)
Moderate complexity; composition of services
- Notebooks parameterizing jobs on Cloud AI Platform (CAIP)
- Traininng and Prediction using CAIP
Intermediate complexity; pipeline orchestration of services
- Notebooks parameterizing pipelines
- Traininng and Prediction using CAIP
- Pipeline orchestration using kubeflow
High complexity; automated orchestration of a CT/CD pipeline
- Pipelines as packaged software
- Custom training pipeline (kubeflow)
- Custom prediction service (cloud functions, cloud run)
- Continuous training, deployment (cloud build)