Google Cloud Professional Machine Learning Engineer Cert Prep

Course details

Earning the Google Cloud Professional Machine Learning Engineer certification confirms that you’re able to build, evaluate, productionize, and optimize AI solutions by using Google Cloud capabilities and knowledge of conventional ML approaches. In this course, Noah Gift prepares you for the certification, starting with an overview of the exam—including the format of the exam, the time it should take, and how and where you can take the exam. Noah then dives into the six sections of the exam, covering what you need to know about: architecting low-code ML solutions; collaborating within and across teams to manage data and models; scaling prototypes into ML models; serving and scaling models; automating and orchestrating ML pipelines; and monitoring ML solutions.

Concepts

Introduction

Course and Google Professional Machine Learning Engineer exam overview
Framing ML problems - Key terminology

Translating Business Challenges into ML Use Cases

Building AI-enabled workflows
Using AI tools to build AI tools
Teaching MLOps at scale with GitHub

Defining ML Problems

Simulations vs. experiment tracking
When to use ML
Supervised vs. unsupervised ML
Optimization
Clustering

Defining Business Success Criteria

Defining business success criteria

Identifying Risks to Feasibility of ML Solutions

MLOps hierarchy of needs
Hidden costs of bespoke systems
Data poisoning

Conclusion - Framing ML Problems

Framing ML problems - Next steps

Introduction - Architecting a ML Solution

Architecting a ML solution - Overview
Architecting a ML solution - Key terminology
Cloud developer workspace advantage

Designing a Reliable, Scalable, and Highly Available ML Solution

What is continuous delivery
Containerized ML microservices
SRE mindset for MLOps
Reproducible workflow
Learn continuous integration

Choosing Appropriate Google Cloud Hardware Components

Selecting heavy vs. light MLOps
Key components of MLOps landscape
Feature store vs. data warehouse
Compute choice

Conclusion - Architecting a ML Solution

Architecting a ML solution - Next steps

Introduction - Designing Data Preparation and Processing Systems

Designing data preparation and processing systems - Overview
Designing data preparation and processing systems - Key terminology
Onboard to GCP

Exploring Data

What is Google Colab
Exploratory data analysis for life expectancy
Data science setup with virtualenv and pip on Windows
Graphing data for exploratory data analysis

Building Data Pipelines

Labeling data
Mechanical Turk labeling
Cleaning up data
Scaling data
BigQuery data pipelines with Colab

Creating Input Features

Feature engineering concepts
Extracting features from public datasets
Exploratory data analysis with Google BigQuery

Conclusion - Designing Data Preparation and Processing Systems

Designing data preparation and processing systems - Next steps

Introduction - Developing ML Models

Developing ML models - Overview
Developing ML models - Key terminology

Building Models

Using TensorFlow Playground
Overfitting vs. underfitting
Selecting the right metrics

Training Models

Training models with TensorFlow and a GPU-enabled Docker
Fine-tuning raw ingredients with Hugging Face
Advantages of transfer learning

Scaling Model Training and Serving

Operationalize microservices
Monitoring and logging with Rust on Google App Engine
Continuous integration using Rust with GitHub Actions
Demo - Unit testing Rust
Demo - GitHub copilot-enabled Rust
Set up a GCP workstation with Python
Demo - Google Cloud Shell
Demo - Google Cloud Editor
Demo - Google CLI SDK
Demo - Google gcloud CLI
Demo - Google App Engine Rust deployment
Demo - Google App Engine Golang

Conclusion - Developing ML Models

Developing ML models - Next steps

Introduction - Automating and Orchestrating ML Pipelines

Automating and orchestrating ML pipelines - Overview
Automating and orchestrating ML pipelines - Key terminology

Designing and Implementing Training Pipelines

Prompt engineering for Google BigQuery with ChatGPT 4
Getting started with Vertex AI
Understanding TPUs
TPUs as a technology transition
Demo - TPU, PyTorch, and MNIST

Implementing Serving Pipelines

TensorFlow serving with a GPU-enabled Docker
A Rust and PyTorch microservice walkthrough
Demo - Rust pretrained PyTorch microservice

Conclusion - Automating and Orchestrating ML Pipelines

Automating and orchestrating ML pipelines - Next steps

Introduction - Monitoring, Optimizing, and Maintaining ML Solutions

Monitoring, optimizing, and maintaining ML solutions - Overview
Monitoring, optimizing, and maintaining ML solutions - Key terminology

ML Solutions

Data drift explained by the naughty child problem
Load testing with Locust
Demo - Auditing via logs
Demo - Logging dashboard
Demo - Cloud web security scanner
Demo - Querying logging output with BigQuery
Demo - Load testing with Rust
Five whys
Using Google Courses
Building a translator with Rust and Hugging Face
Using PyTorch and Rust for stable diffusion
Using Rust with PyTorch
Building a CUDA GPU stress test

Conclusion

Next steps