You are tasked with building an ML pipeline on Google Cloud for data processing, model training, and model deployment. Your solution involves different Google Cloud services such as Cloud Storage and Vertex AI. Each task in your pipeline has been individually coded and you anticipate a high frequency of new data files being added to your Cloud Storage bucket. Now, you need an orchestration layer to manage these tasks efficiently. This orchestration pipeline should only run when new files are detected in the Cloud Storage bucket to avoid unnecessary computations and minimize costs. What approach should you take to achieve this?