Course 825:
MLOps on Google Cloud

(2 days)

 

Course Description

Machine Learning can be used to create predictive algorithms using your data and today’s extremely powerful computers. However, getting these ML algorithms from the research lab into production can be challenging. In this course, you will learn to use the power of Google Cloud and its suite of data analytics and ML tools to more easily process your data, train your models, deploy your models, and manage them in production.

Learning Objectives

  • Simplify and automate machine learning operations using Google Cloud tools
  • Learn the basics of Machine Learning
  • Implement the steps required to build a production ML pipeline
  • Build batch and streaming data pipelines with Google Cloud Dataflow
  • Perform Feature Engineering using Dataflow and BigQuery
  • Train and validate ML models using BigQuery ML, Auto ML, and TensorFlow
  • Deploy models using Google Cloud automation
  • Orchestrate ML Models using Cloud Composer and Kubeflow
  • Leverage Vertex AI to manage ML datasets, models, experiments, and deployment

Who Should Attend

Data Engineers, developers, and others who are working on machine learning projects and want to learn how to use Google Cloud tools to automate the training and deployment of their ML models.

Hands-On Activities

This course includes hands-on labs that reinforce your learning and provide a template for getting started on your real-world use cases.


Course Outline

1. Machine Learning

  • ML Basics
    • ML Examples
    • Models
    • Examples
    • Features 
    • Labels
    • Splitting Training and Validation Data
    • Training
    • Validation
    • Hyperparameters
    • Prediction
  • Linear Regression Models
    • Linear Regression
    • Example
    • Validating Linear Regression Models
    • RMSE
    • Exercise: Building a Linear Regression Model 
  • Classification Models
    • Classification
    • Example
    • Validating Classification Models
    • Accuracy 
    • Precision
    • Recall
    • Area Under Receiver Operator Curve (AUROC)
    • Exercise: Building a Classification Model 
  • Neural Networks
    • Neural Networks
    • Neurons
    • Layers
    • Example
    • Exercise: Building a Neural Network 

2. MLOps 

  • ML Steps
    • Data Collection
    • Feature Engineering
    • Training
    • Validation
    • Putting Models Into Production
  • ML Pipelines
    • Batch ML Pipelines
    • Online Prediction Models
    • On-device Prediction Models
    • Streaming ML Pipelines

3. Creating Data Pipelines with Dataflow

  • Dataflow Basics
  • Apache Beam
    • Creating a Pipeline
    • Pipeline Basics
    • PCollections
    • Transforms
    • I/O
    • Dataflow Templates
    • Exercise: Build a Simple Apache Beam Pipeline 
  • Batch Dataflow Pipelines
    • Example
    • Exercise: Building a Batch Pipeline using a DataFlow Template 
  • Streaming Dataflow Pipelines
    • Example
    • Exercise: Building a Streaming Pipeline using a DataFlow Template 

 4. Feature Engineering

  • Feature Engineering in Dataflow
    • Writing Transforms
    • ParDo
    • Side Inputs
    • Dataflow SQL
    • Exercise: Feature Engineering with Dataflow  
  • Feature Engineering is BigQuery
    • Tuning Features with SQL
    • Joins
    • Filters 
    • User-Defined Functions
    • Saving Results to Tables
    • Exporting Tables
    • Exercise: Feature Engineering with BigQuery

5. Training and Validating ML Models on Google Cloud

  • BigQuery ML
    • Training ML Models with SQL
    • Model Types
    • Options
    • Splitting the Data
    • Training
    • Validation
    • Prediction
    • Exercise: Programming the Iris Classifier in BigQueryML
  • AutoML
    • Tables
    • Vision
    • Natural Language
    • Video
    • Creating Datasets
    • Labeling
    • Training
    • Validation
    • Exercise: Using AutoML Vision 
  • TensorFlow
    • TensorFlow Example
    • ML Notebooks
    • Training Custom Models
    • AI Platform
    • Exercise: Training a TensorFlow Model on AI Platform

6. Deploying ML Models in Google Cloud

  • Deploying BigQuery ML Models
  • Deploying AutoML Models
  • Deploying Custom Models

7. Orchestrating ML Pipelines

  • Composer
  • Kubeflow

8. Vertex AI

  • Vertex AI Basis
  • Managing Training Data 
    • Datasets
    • Features
    • Labeling Tasks
  • Managing Models
    • Training 
    • Deploying Models
    • Endpoints
    • Batch Predictions

 

Please Contact Your ROI Representative to Discuss Course Tailoring!