Amazon SageMaker

 

The Amazon SageMaker course aims to equip learners with a comprehensive understanding of machine learning workflows using AWS by teaching them how to prepare data, build and train models, perform hyperparameter tuning, and deploy scalable ML solutions. Students will learn to work with SageMaker Studio, Data Wrangler, Ground Truth, and built-in ML algorithms, as well as deploy real-time and batch inference endpoints with best practices for monitoring, security, and cost optimization. The course also covers MLOps using SageMaker Pipelines, integration with key AWS services, advanced features like JumpStart, Canvas, Clarify, and Neo, and concludes with a real-world ML project to prepare learners for AWS Machine Learning roles and certification.

Course Content

​

Module 1: Introduction to AWS & SageMaker

• What is Machine Learning? (Supervised, Unsupervised, Reinforcement Learning)

• Challenges in traditional ML workflows

• Introduction to AWS AI/ML ecosystem

• What is Amazon SageMaker?

• Key benefits of SageMaker for data scientists & engineers

• Use cases: NLP, computer vision, forecasting, fraud detection

Module 2: SageMaker Architecture & Key Components

• SageMaker Studio & Studio Lab

• SageMaker Notebook Instances

• SageMaker Compute options (CPU, GPU, Accelerators)

• SageMaker Domain, Users, IAM roles

• Storage: S3, EFS, FSx integration

• Security: IAM, VPC, Encryption

Module 3: Data Preparation with SageMaker

• Importing datasets from S3

• Using SageMaker Data Wrangler

• Feature engineering and transformation

• Data labeling with SageMaker Ground Truth

• Data quality monitoring

Module 4: Training Models in SageMaker

• Built-in algorithms

• Frameworks: TensorFlow, PyTorch, Scikit-learn

• SageMaker Estimators

• Hyperparameters and optimization

• Distributed training

• Spot instances for training cost optimization

Module 5: Model Tuning & Optimization

• Automatic Model Tuning (Hyperparameter Tuning Jobs)

• Warm-start training

• Metrics & logs

• Debugger & Profiler

• Feature importance tools

Module 6: Model Deployment

• Inference concepts (Real-time, Batch, Asynchronous)

• Setting up endpoints

• Multi-model endpoints

• Serverless inference

• A/B testing & blue/green deployments

• Auto-scaling

• Cost optimization & pricing

Module 7: MLOps with SageMaker

• Introduction to MLOps

• SageMaker Pipelines

• CI/CD integration with GitHub, CodePipeline

• Model registry

• Automating retraining workflows

• Monitoring model drift