Machine Learning in Production

Description

In this course, you will learn the best practices for managing machine learning experiments and models with MLflow. There are two main components in this course: (i) using MLflow to track the machine learning lifecycle, package models for deployment, and manage model versions (ii) examining various production issues, different deployment paradigms, and post-production concerns. By the end of this course, you will have built an end-to-end pipeline to log, deploy, and monitor machine learning models.

This course will prepare you to take the Databricks Certified Machine Learning Professional exam.

Duration

1 full day or 2 half days

Objectives

• Track, version, and manage machine learning experiments
• Leverage Databricks Feature Store for reproducible data management
• Implement strategies for deploying models for batch, streaming, and real-time
• Build monitoring solutions, including drift detection

Prerequisites

• Intermediate experience with Python and pandas
• Working knowledge of machine learning and data science (scikit-learn, TensorFlow, etc.)
• Familiarity with Apache Spark

Outline

Day 1

• ML in production overview
• Data management with Delta and Databricks Feature Store
• Experiment tracking and versioning with MLflow Tracking
• Model management with MLflow Models and Model Registry
• Automated testing with webhooks
• Deployment paradigms
• Monitoring and CI/CD

If you have any questions, please refer to our Frequently Asked Questions page.