MLOps Bootcamp: Build Real-World AI Infrastructure

Overview

MLOps is the discipline that combines AI/Machine Learning, Data Engineering, and DevOps to reliably deploy, monitor, and maintain AI models. You'll learn to go beyond model training and learn how real AI systems operate in production by connecting AWS services into a modern MLOps stack that take an AI model from experimentation to a live, continuously improving, production-ready AI application that scales.

Understand the full MLOps lifecycle from training to retraining
Fine-tune Vision Transformers with Hugging Face on AWS
Prepare S3 datasets for cloud-based model training
Deploy real-time AWS endpoints for live predictions
Monitor model confidence and activity with CloudWatch
Build AWS Pipelines for automated ML workflows
Connect Lambda, API Gateway, S3 and Bedrock
Retrain models automatically and gate production deployments

Syllabus

Introduction

Introduction
Exercise: Meet Your Classmates and Instructor
What We Are Building
Demo: How The Retrain Pipeline Gets Kicked Off
Course Resources
Understanding Your Video Player
Set Your Learning Streak Goal

Understanding AWS, IAM roles, Domains and Notebooks

Signing in to AWS
Understanding IAM Roles and Permissions
Setting Up SageMaker Domain
Understanding SageMaker Notebooks
Let's Have Some Fun (+ More Resources)

Installing Required Packages and Dependencies to our Sagemaker Notebook

Installing Sagemaker Versions
Importing Libraries and Tools
Important Reminder to Shut Down Your AWS Server
Unlimited Updates

Accessing our Dataset and doing Exploratory Data Analysis

Downloading the Data From HuggingFace
Addressing Rate Limit Errors
Exploratory Data Analysis
Uploading Dataset to S3
Verifying the Dataset in S3
Implement a New Life System

Setting up our AWS Sagemaker Training Job with Vision Transformers

Configuring Our Vision Transformer Train Setup
Creating Helper Functions and Evaluations for the Train Script
Creating Main Train Function in the Train Script
Setting Up Model Configurations for Training
Getting Metrics During Training
Saving Model Artifacts
Setting Up SageMaker Estimator

Training our Vision Transformer on Nvidia GPUs in AWS

Starting the Training Job on a GPU Server
Debugging Training Errors
Evaluating the Training Job
Course Check-In

Deploying our Model to Sagemaker and Setting up Monitoring Dashboards

Creating the Inference Script - Part 1
Creating the Inference Script - Part 2
Creating the Inference Script - Part 3
Deploying a Live Endpoint
Testing our Endpoint and our Data Capture Configuration
Setting Up CloudWatch Dashboards for Monitoring our Pipeline
Shutting Down the Notebook Server

Creating our Sagemaker Retrain Pipeline

Understanding the Re-Training Pipeline
Creating The Endpoint Update Lambda Function for the SageMaker Pipeline
Endpoint Configuration for SageMaker Pipeline and A_B Test Setup
Finishing The Lambda Deploy Model Function
Comparing Model Metrics in the Re-Train Pipeline
Getting the Candidate Model_s Metrics
Brief Explanation on the Previous Video
Compare Metrics Function
Saving Comparison Metrics to the SageMaker Directory
Importing Modules for the SageMaker Pipeline
Creating Pipeline Definitions
Setting Up the Processing Steps in the SageMaker Pipeline
Adding The Comparison Step to the SageMaker Pipeline
Adding the Conditional Lambda Deployment to the SageMaker Pipeline
Clarification on Condition Step
Exercise: Imposter Syndrome

Connecting the Sagemaker Pipeline with other AWS Services

Understanding The Lambda Function that Connects API Gateway with our SageMaker Pipeline
Setting Up The Connection Between API Gateway, SageMaker Endpoint, and the Pipeline - Part 1
Setting Up The Connection Between API Gateway, SageMaker Endpoint, and the Pipeline - Part 2
Clarification Reagrding the List Pending Examples Function
Setting Up The Connection Between API Gateway, SageMaker Endpoint, and the Pipeline - Part 3
Setting Up The Connection Between API Gateway, SageMaker Endpoint, and the Pipeline - Part 4
Setting Up The Connection Between API Gateway, SageMaker Endpoint, and the Pipeline - Part 5
Updating Our Train Function to Have the Retrain Method
Setting Up API Gateway and the Lambda Integration
Importing Pipeline Definition File to Our SageMaker Notebook
Creating our SageMaker Pipeline and Visualizing the Graph_udemy_fixed
Setting Up Deployment Notebook for Professional Deployment
Testing API Gateway and Our CloudWatch Dashboard_udemy_fixed
Checking the Retraining Pipeline - Part 1
Updating the SageMaker Pipeline with the Appropriate GPU Server
Updating Lambda Function with Appropriate Permissions