Production Machine Learning Systems

In this course, we dive into the components and best practices of building high-performing ML systems in production environments. We cover some of the most common considerations behind building these systems, e.g. static training, dynamic training, static inference, dynamic inference, distributed TensorFlow, and TPUs. This course is devoted to exploring the characteristics that make for a good ML system beyond its ability to make good predictions.

• Introduction to Advanced Machine Learning on Google Cloud
• This module previews the topics covered in the course and how to use Qwiklabs to complete each of your labs using Google Cloud.
• Architecting Production ML Systems
• This module explores what else a production ML system needs to do and how to meet those needs. You review how to make important, high-level, design decisions around training and model serving need to make in order to get the right performance profile for your model.
• Designing Adaptable ML Systems
• In this module, you learn how to recognize the ways that our model is dependent on our data, make cost-conscious engineering decisions, know when to roll back our models to earlier versions, debug the causes of observed model behavior and implement a pipeline that is immune to one type of dependency.
• Designing High-Performance ML Systems
• In this module, you identify performance considerations for machine learning models. Machine learning models are not all identical. For some models, you focus on improving I/O performance, and on others, you focus on squeezing out more computational speed.
• Building Hybrid ML Systems
• Understand the tools and systems available and when to leverage hybrid machine learning models.
• Summary
• This module reviews what you learned in this course.