Having now walked you through the under-the-hood details of how Kubernetes, Docker, and AWS can be used to throw as many resources at your model as your wallet will allow, there are a number of steps you can take to customize these examples to your use case or take your level of knowledge even further:

• Integrate this approach into your CI or CD tool (Bamboo, CircleCI, Puppet, and so on)
• Integrate Pachyderm into your Docker, Kubernetes, or AWS solution
• Experiment with the parameter server to do things such as distributed gradient descent and further optimize your model pipeline