Production Machine Learning Systems
By Google Cloud
In this course, we will dive into the components and best practices of a high-performing ML system in production environments.
Course info
Level
Advanced

Updated
Sep 17, 2019

Duration
3h 18m

Table of contents
Welcome to the course
Architecting Production ML Systems
Introduction
2m
The Components of an ML System
2m
The Components of an ML System:Data Analysis and Validation
4m
The Components of an ML System:Data Transformation + Trainer
2m
The Components of an ML System:Tuner + Model Evaluation and Validation
2m
The Components of an ML System:Serving
1m
The Components of an ML System:Orchestration + Workflow
4m
The Components of an ML System:Integrated Frontend + Storage
2m
Training Design Decisions
5m
Serving Design Decisions
5m
Lab Intro:Serving on Cloud AI Platform
2m
Serving on Cloud AI Platform
0m
Lab Solution:Serving on Cloud AI Platform
4m
Designing from Scratch
3m
Ingesting data for Cloud-based analytics and ML
Introduction
5m
Data On-Premise
3m
Large Datasets
4m
Data on Other Clouds
2m
Existing Databases
2m
Demo:Load data into BigQuery
5m
Demo:Automatic ETL Pipelines into GCP
4m
Designing Adaptable ML systems
Introduction
4m
Adapting to Data
3m
Changing Distributions
3m
Exercise:Adapting to Data
2m
Right and Wrong Decisions
3m
System Failure
2m
Mitigating Training-Serving Skew through Design
1m
Lab Intro:Serving ML Predictions in batch and real-time
2m
Serving ML Predictions in batch and real-time
0m
Lab Solution:Serving ML Predictions in batch and real-time
9m
Debugging a Production Model
4m
Summary
1m
Designing High-performance ML systems
Introduction
1m
Training
6m
Predictions
3m
Why distributed training?
5m
Distributed training architectures
6m
Faster input pipelines
3m
Native TensorFlow Operations
3m
TensorFlow Records
1m
Parallel pipelines
6m
Data parallelism with All Reduce
5m
Parameter Server Approach
3m
Inference
4m
Hybrid ML systems
Introduction
5m
Machine Learning on Hybrid Cloud
5m
KubeFlow
3m
Demo:KubeFlow
22m
Kubeflow: End to End
0m
Embedded Models
3m
TensorFlow Lite
2m
Optimizing for Mobile
6m
Summary
2m
Course Summary
Description
Course info
Level
Advanced

Updated
Sep 17, 2019

Duration
3h 18m

Description
In this course, we will dive into the components and best practices of a high-performing ML system in production environments.
About the author