MLOps Engineering on AWS
Please not that this is not a prep course for the “AWS Certified Machine Learning Engineer - Associate”, “AWS Certified Machine Learning - Specialty” or any other certification!
This document contains links mentioned in the “MLOps Engineering on AWS” course.
Module 0: Course Overview
Full name: MLOps Engineering on AWS
Product code: 200-MLPOPS-20
These links are based on the course version 2.0.4.
The official course material: eVantage Bookshelf (VitalSource)
Additional material: AWS Skill Builder
Module 1: Introduction to MLOps
Websites
- MLOps Principles
- Hidden Technical Debt in Machine Learning Systems
- Amazon SageMaker Studio Classic
- Secure Data Science With Amazon SageMaker Studio Reference Architecture
- Amazon EMR
- Amazon Athena
- AWS Glue
- AWS CodeCommit
Module 2: Initial MLOps /Experimentation Environments in SageMaker Studio
Websites
- Use Amazon SageMaker Studio Classic Notebooks
- How Are Amazon SageMaker Studio Classic Notebooks Different from Notebook Instances?
- Use Docker containers to train and deploy models
- Use Amazon SageMaker built-in algorithms or pretrained models
- Use Pre-built SageMaker Docker images
- Machine Learning Frameworks and Languages
- Available Deep Learning Containers Images
- What are AWS Deep Learning Containers?
- Create a container with your own algorithms and models
- Adapting your own Docker container to work with SageMaker
- Build container images from your Studio notebooks
- SageMaker Environment Variables and Default Paths for Training Storage Locations
- Adapting your own Docker container to work with SageMaker
- Bring your own pre-trained MXNet or TensorFlow models into Amazon SageMaker
- Use Docker containers to train and deploy models
- Dive deep into Amazon SageMaker Studio Classis Notebooks architecture
- Install External Libraries and Kernels in Amazon SageMaker Studio Classic
- Quick setup to Amazon SageMaker
- Automate a centralized deployment of Amazon SageMaker Studio with AWS Service Catalog
- SageMaker Studio Permissions Required to Use Projects
- Create Amazon SageMaker projects with image building CI/CD pipelines
- Train, deploy, and evaluate pretrained models with SageMaker JumpStart
- Collaborate with shared spaces
- Persona reference
- ML activity reference
- Configuring Amazon SageMaker Studio for teams and groups with complete resource isolation
- Best Practicies for Developing Life-Cycle Configuration Scripts for SageMaker Studio
- Install External Libraries and Kernels in Amazon SageMaker Studio Classic
- SageMaker Studio Classic Lifecycle Configuration Samples
- Verify lifecycle configuration process from CloudWatch Logs
- SageMaker Studio Classic Lifecycle Configuration Samples
- Customize Amazon SageMaker Studio using Lifecycle Configurations
- Using CodeWhisperer and CodeGuru extensions with SageMaker
- Amazon SageMaker Studio and SageMaker Notebook Instance now come with JupyterLab 3 notebooks to boost developer productivity
- MLOps foundation roadmap for enterprises with Amazon SageMaker
- AWS Solutions Library
Module 3: Repeatable MLOps / Repositories
Websites
- Using versioning in S3 buckets
- Lake Formation tag-based access control
- Amazon SageMaker Feature Store
- Create, store, and share features with Feature Store
- Amazon SageMaker Data Wrangler
- Batch ingestion with Amazon SageMaker Feature Store Spark
- Feature Store Feature Processor SDK
- Unlock ML insights using the Amazon SageMaker Feature Store Feature Processor
- Create and Manage SageMaker Pipelines
- Register and Deploy Models with Model Registry
- Register a Model Version
- Add a Git Repository to Your Amazon SageMaker Account
- Authentication and access control for AWS CodeCommit
Module 4: Repeatable MLOps / Orchestration
Websites
- MLOps Workload Orchestrator on AWS
- Use the XGBoost algorithm with Amazon SageMaker
- Orchestrate Jobs to Train and Evaluate Models with Amazon SageMaker Pipelines
- IAM: Pass an IAM role to a specific AWS service
- IAM roles
- Using the SageMaker Python SDK
- Pipeline Parameters
- Pass Data Between Steps
- Caching Pipeline Steps
- Selective execution of pipeline steps
- Retry Policy for Pipeline Steps
- Create and Manage SageMaker Pipelines
- Amazon SageMaker pricing
- Deploy models with Amazon SageMaker Serverless Inference
- Asynchronous inference
- Use batch transform to run inference with Amazon SageMaker
- Amazon SageMaker pricing
- Best practices for deploying models on SageMaker Hosting Services
- Amazon SageMaker Model Building Pipelines steps
- Kubernetes Orchestration
- Manage SageMaker with Step Functions
- Train a Machine Learning Model
- Tune a Machine Learning Model
- Preprocess data and train a machine learning model
- What Is AWS Service Catalog?
- Use SageMaker-Provided Project Templates
- SageMaker Studio Permissions Required to Use Projects
- AWS Managed Policies for SageMaker projects and JumpStart
- Orchestrate XGBoost ML Pipelines with Amazon Managed Workflows for Apache Airflow
- Managing your machine learning lifecycle with MLflow and Amazon SageMaker
- Securing MLflow in AWS: Fine-grained access control with AWS native services
- Define customized permissions in minutes with Amazon SageMaker Role Manager
- New ML Governance Tools for Amazon SageMaker – Simplify Access Control and Enhance Transparency Over Your ML Projects
- Improve governance of your machine learning models with Amazon SageMaker
- Amazon SageMaker controls
- Securing Amazon SageMaker Studio connectivity using a private VPC
- Connect to Amazon SageMaker resources from within a VPC
Module 5: Reliable MLOps / Scaling and testing
Websites
- Host models along with pre-processing logic as serial inference pipeline behind one endpoint
- Use a multi-container endpoint with direct invocation
- Configuring autoscaling inference endpoints in Amazon SageMaker
- Step scaling policies for Application Auto Scaling
- Organizing Your AWS Environment Using Multiple Accounts
- Bootstrapping multiple AWS accounts for AWS CDK using CloudFormation StackSets
- Create Custom Project Templates
- Benefits of using multiple AWS accounts
- Operating model
- AWS Well-Architected
- Machine Learning Lens
- Testing phase
- Shadow tests
- Amazon Sagemaker API Reference: InvokeEndpoint
- Amazon Sagemaker API Reference: CreateEndpoint
- Production variants
- CloudWatch Metrics
- Update the Approval Status of a Model
- Perform Shadow Tests to Compare Inference Performance Between ML Model Variants
- Monitor Amazon SageMaker with Amazon CloudWatch
- Getting started with CodeWhisperer
Videos
Module 6: Reliable MLOps / Monitoring
Websites
- Monitor Amazon SageMaker with Amazon CloudWatch
- Visual explanations of core machine learning concepts
- Transform responsible AI from theory into practice
- Amazon CloudWatch
- Amazon SageMaker Studio
- Model quality metrics and Amazon CloudWatch monitoring
- SageMaker Model Monitor - visualizing monitoring results
- Schedule monitoring jobs
- Monitor Bias Drift for Models in Production
- Extend model lineage to include ML features using Amazon SageMaker Feature Store
- Amazon SageMaker pricing
- Monitor Bias Drift for Models in Production
- Measure Pre-training Bias
- Measure Post-training Data and Model Bias
- Schedule model quality monitoring jobs
- Querying Lineage Entities
Module 7: Course Wrap-up
Websites
- MLOps foundation roadmap for enterprises with Amazon SageMaker
- Create Custom Project Templates
- Custom Project Templates in SageMaker
- Amazon SageMaker for MLOps
- MLOps Workload Orchestrator on AWS
- AWS Certification exam preparation
- AWS Certification
- AWS Recertification
- All AWS Certification exams
- Schedule an Exam
- AWS Skill Builder subscription
- AWS Training and Certification
- AWS Online Tech Talks
- AWS Ramp-Up Guides