Summary

In this chapter, you learned about the core architecture components of a typical ML platform and their capabilities. We also discussed various open source technologies such as Kubeflow, MLflow, TensorFlow Serving, Seldon Core, Apache Airflow, and Kubeflow Pipelines. You have also built a data science environment using Kubeflow notebooks, tracked experiments and models using MLflow, and deployed your model using Seldon Core. And finally, you learned how to automate multiple ML workflow steps using Kubeflow Pipelines, including data processing, model training, and model deployment. While these open source technologies provide features for building potentially sophisticated ML platforms, it still takes significant engineering effort and know-how to build and maintain such environments, especially for large-scale ML platforms. In the next chapter, we will start looking into fully managed, purpose-built ML solutions for building and operating ML environments.