ML Production Pipelines: Building and Deploying a Classification Model
Explore the intricacies of productionalizing a machine learning pipeline in this 27-minute talk from Databricks. Dive into the deployment of a classification model, showcasing how Python, Databricks, and MLflow can be integrated to create robust production pipelines. Learn about the dual pipeline approach - training and prediction - and how S3 Bucket serves as the data source. Discover the process of training various models, registering them in MLflow, and storing metrics and hyperparameters. Understand how Grid Search is utilized to select the best model for production, and explore deployment options using Flask or UDF for batch processing. Gain insights into packaging the entire project as a library for easy installation and configuration. From introduction to conclusion, cover topics such as development, training, code structure, architecture, MLflow integration, and Databricks implementation.