 It's great to work and train small models in your local environment. Take as an example this notebook using the newly released Keras Core with Jack's backend. Here we are training a model to classify flowers using the popular Flowers dataset. But what if you want your friends or customers to try your model and provide feedback? Let's take this one step further. What about selecting an image online? Let's say this tulip and uploading it to a web application and getting predictions. That's way easier and shareable. Or what if the model needs retraining? Can we automate this process? Let's check it out. Let's take the neural network responsible for classification and add more epochs to the training process. For illustration purposes, let's set it to 20. Commit this change. And we see here that this hyperparameter change automatically started a pipeline that is not only building the container with the neural network code, but ensuring that no cold vulnerabilities are being introduced to the codebase. All of this before triggering the retraining of the model. So far, the pipeline is successful. Let's take a look at the training job. Right now, GitLab is interacting with the Vertex AI platform. It is executing a custom training job using the container we built with the neural network in the first step of the CI pipeline. I'm going to click here and see more details about the training pipeline. All right, this shows me the GPU hardware provision to run the model retraining and I can see where the train model is going to be pushed as an artifact. All right, the job finished. And I want to highlight some steps that automatically happened here. With the model training done, a new Flowers classifier model was created. In addition to that, an endpoint to interact with the Flowers model was also automatically created and deployed. All of this with security in mind. In the upcoming demos, we will learn how to fill this gap deeper into the steps to automate training and deployment of this Flowers application and take our prototypes to production faster using DevSecOps.