AWS: Model Training , Optimization & Deployment

Whizlabs via Coursera

Go to class Write review

Overview

Coursera Flash Sale

40% Off Coursera Plus for 3 Months!

Grab it

AWS: Model Training, Optimization & Deployment is the third course in the Exam Prep (MLA-C01): AWS Certified Machine Learning Engineer – Associate Specialization. This course is designed to equip learners with the skills to train, optimize, and deploy machine learning models efficiently using AWS services. Learners begin by exploring popular algorithms such as Linear Learner, XGBoost, LightGBM, and k-Nearest Neighbors (k-NN), and understand their use cases in classification and regression tasks.You’ll then dive into the model training process, learning how to configure key parameters like epochs, batch size, and steps for optimized performance. Then the learners will begin by exploring SageMaker Model Debugger and SageMaker Experiments, which help monitor training jobs and compare experiment results efficiently.You’ll then dive into cross-validation techniques and learn how to apply hyperparameter tuning using both random search and Bayesian optimization methods to improve model accuracy. Finally by exploring compute options such as Amazon ECS, Amazon EKS, and AWS Lambda, followed by infrastructure management with AWS CloudFormation.You’ll learn how to implement auto scaling policies for ML workloads and choose the right SageMaker compute instance types (CPU vs. GPU) for different deployment scenarios. This course is divided into three comprehensive modules, each containing targeted lessons and practical demonstrations. Learners will benefit from approximately 3.5 to 4 hours of expert-led video content, featuring real-world use cases and hands-on walkthroughs using AWS tools. Every module includes Graded and Ungraded Quizzes to assess conceptual understanding and application. Module 1: Model Training, Algorithms & Inference Techniques Module 2: Model Optimization, Evaluation & Tuning with SageMaker Module 3: Scalable Infrastructure & Automated ML Deployment on AWS By the end of this course, learners will be able to: Compare real-time and batch inference approaches to determine the best strategy for model deployment. Apply model optimization techniques such as hyperparameter tuning Understand and select appropriate inference strategies for deployment Explore AWS compute and orchestration services like ECS, EKS, Lambda, and CloudFormation for ML deployment. This course is ideal for ML practitioners, data scientists, and cloud developers who are looking to scale their ML workflows and gain hands-on experience with advanced features of Amazon SageMaker. It is also designed for learners preparing for the AWS Certified Machine Learning Engineer – Associate (MLA-C01) exam, focusing on the model training and deployment aspects of the certification.

Syllabus

Model Training, Algorithms & Inference Techniques

Welcome to Week 1 of the AWS: Model Training, Optimization & Deployment course. This week, you’ll focus on building machine learning models using Amazon SageMaker’s built-in algorithms. We’ll begin by exploring popular algorithms such as Linear Learner, XGBoost, LightGBM, and k-Nearest Neighbors (k-NN), and understand their use cases in classification and regression tasks. You’ll then dive into the model training process, learning how to configure key parameters like epochs, batch size, and steps for optimized performance. Through hands-on demos, you’ll practice training models, splitting datasets into train-test sets, and preparing them for evaluation. We’ll conclude the week by comparing real-time vs. batch inference, helping you understand how to choose the appropriate inference strategy based on your workload and deployment needs.

Model Optimization, Evaluation & Tuning with SageMaker

Welcome to Week 2 of the AWS: Model Training, Optimization & Deployment course. This week, you'll focus on optimizing and managing machine learning models to ensure high performance and reliability in production environments. We'll begin by exploring SageMaker Model Debugger and SageMaker Experiments, which help monitor training jobs and compare experiment results efficiently. You’ll then dive into cross-validation techniques and learn how to apply hyperparameter tuning using both random search and Bayesian optimization methods to improve model accuracy. We’ll also cover model ensembling techniques, such as stacking and boosting, to combine multiple models for better predictive power. By the end of the week, you’ll learn how to manage model versions using SageMaker Model Registry, apply automatic model tuning, and implement strategies to detect and prevent overfitting or underfitting for building robust ML solutions.

Scalable Infrastructure & Automated ML Deployment on AWS

Welcome to Week 3 of the AWS: Model Training, Optimization & Deployment course. This week, you’ll focus on deploying machine learning models efficiently using scalable infrastructure and automation tools on AWS. We’ll begin by exploring compute options such as Amazon ECS, Amazon EKS, and AWS Lambda, followed by infrastructure management with AWS CloudFormation. You’ll learn how to implement auto scaling policies for ML workloads and choose the right SageMaker compute instance types (CPU vs. GPU) for different deployment scenarios. We'll also cover SageMaker Endpoint types, including serverless, asynchronous, and multi-model endpoints, to help you deliver predictions at scale. Finally, you’ll dive into workflow orchestration using Apache Airflow and SageMaker Pipelines, and understand the role of CI/CD principles in automating and streamlining ML deployments.