You are developing an ML model for real-time video processing that involves slicing frames from a video feed and creating bounding boxes around specific objects. Your goal is to automate the entire training pipeline on Google Cloud. This involves: ingestion and preprocessing of data stored in Cloud Storage, training and hyperparameter tuning of the object detection model using Vertex AI jobs, and deploying the model to an endpoint. To achieve this, you want to orchestrate the end-to-end pipeline with minimal cluster management. Given these requirements, what approach should you use?