How to deploy an on-demand GPU heavy VM instance on GCP in a nutshell?

Question

I have a GPU-heavy image processing workload (A Python script that takes in a few hundred images) that I need to deploy on GCP. What is the easiest way to do this? Do I need