Amazon SageMaker Machine Learning Training Course

Course overview

Teams often have data in Amazon S3 and promising machine learning use cases, but struggle to turn experiments into governed, repeatable services. Analysts and engineers may be able to train a notebook model, yet lack a clear process for selecting algorithms, managing feature data, evaluating model quality, deploying endpoints, and monitoring operational performance. This course addresses that gap using Amazon SageMaker as the working environment for the full machine learning lifecycle.

Participants build practical capability across SageMaker Studio, notebook environments, data preparation, built-in algorithms, training jobs, hyperparameter tuning, model evaluation, endpoint deployment, and monitoring. They work with Amazon S3 datasets and the SageMaker Python SDK to create reproducible workflows rather than one-off notebook experiments. The course also covers IAM considerations, experiment tracking, model registry practices, inference options, cost-aware compute selection, and the metrics needed to assess classification and regression models.

Instruction combines guided demonstrations with lab-based work on a realistic business prediction case. Each participant configures a SageMaker workspace, prepares a dataset, trains and compares models, registers an approved model version, deploys an inference endpoint, and reviews monitoring signals. They leave with a documented SageMaker machine learning implementation blueprint: an architecture diagram, model evaluation record, deployment decision, and practical next-step plan that can be adapted to a workplace use case.

The course is suited to data professionals, cloud practitioners, software engineers, and technical product staff who need to contribute credibly to machine learning delivery on AWS. It is particularly valuable for organisations moving from proof-of-concept models to controlled, supportable production workflows.

Course objectives

By the end of this course, participants will be able to:

• Configure an Amazon SageMaker Studio domain, user profile, storage access, and execution role for machine learning work.
• Prepare tabular training data from Amazon S3 using SageMaker Data Wrangler transformations and data-quality checks.
• Build reproducible training workflows with SageMaker notebooks and the SageMaker Python SDK.
• Train and evaluate classification and regression models using SageMaker built-in algorithms and managed training jobs.
• Apply hyperparameter tuning jobs and validation metrics to select a defensible model candidate.
• Track experiments, register approved model versions, and document model approval criteria in SageMaker Model Registry.
• Deploy a model to a real-time SageMaker endpoint and test inference requests with representative payloads.
• Design a SageMaker implementation blueprint covering architecture, monitoring, security controls, ownership, and cost considerations.

Benefits of attending

Target competencies

Who should attend

• Data Analysts — who need to turn prepared business data into evaluated machine learning models on AWS.
• Data Scientists — who need a managed environment for repeatable training, experiment tracking, and model deployment.
• Machine Learning Engineers — who need to operationalise SageMaker training and inference workflows.
• Cloud Engineers — who support AWS platforms and must configure secure, cost-conscious SageMaker environments.
• Software Engineers — who integrate prediction endpoints into applications and services.
• Technical Product Managers — who scope machine learning use cases and need to assess delivery, governance, and operating requirements.

Requirements and prerequisites

Participants should be comfortable working with files and datasets, using a web-based cloud console, and reading basic Python code. Familiarity with tabular data concepts such as rows, columns, missing values, training and test sets, and simple evaluation measures is helpful. Prior AWS experience is useful but not essential; the course explains the required Amazon S3, IAM, and SageMaker concepts in context. Participants do not need previous data science employment, advanced statistics, deep learning experience, or prior SageMaker certification. A complete beginner should expect a technically practical week and should review basic Python syntax before attending.

Training methodology

The five days alternate short instructor-led explanations with guided SageMaker labs in which participants work through one tabular prediction use case. Demonstrations show the purpose and configuration of each AWS service before participants implement it in Studio and with the SageMaker Python SDK. Small-group reviews are used to challenge feature choices, evaluation metrics, deployment options, and governance decisions. Daily exercises produce working artefacts, culminating in an individual implementation blueprint that connects a trained model to a realistic AWS operating model.

Course outline

Day 1: SageMaker foundations and machine learning workflow

• Machine learning lifecycle from business problem to monitored inference
• Amazon SageMaker service components and deployment patterns
• SageMaker Studio domains, user profiles, and shared project workspaces
• IAM execution roles, least-privilege access, and Amazon S3 data permissions
• SageMaker notebook instances and Studio notebook environments
• Training, validation, and test dataset design
• Classification and regression metrics for business model selection

Workshop: Participants configure a SageMaker Studio workspace and produce a use-case canvas defining the prediction target, data sources, stakeholders, and success metric.

Day 2: Data preparation and feature engineering

• Loading CSV and Parquet datasets from Amazon S3
• Data profiling and missing-value analysis in SageMaker Data Wrangler
• Data Wrangler transformation flows and export options
• Categorical encoding, scaling, and date-derived feature creation
• Data leakage detection and feature selection principles
• Train-validation-test splitting and stratified sampling
• Feature Store concepts and reusable feature governance

Workshop: Participants use SageMaker Data Wrangler to profile a business dataset, create transformations, and export a versioned training dataset with a data-preparation record.

Day 3: Model training, experimentation, and evaluation

• SageMaker Python SDK estimators, sessions, and training job configuration
• SageMaker built-in algorithms including XGBoost and linear learner
• Managed training jobs, input channels, output artefacts, and logs
• SageMaker Experiments for run comparison and parameter tracking
• Hyperparameter tuning jobs and search-range design
• Confusion matrices, precision, recall, F1 score, RMSE, and MAE
• Bias, overfitting, and model-selection decisions

Workshop: Participants train two candidate models, run a hyperparameter tuning job, and produce a model comparison table recommending a preferred version.

Day 4: Model deployment and production integration

• SageMaker model packages, inference containers, and model artefacts
• SageMaker Model Registry groups, versions, and approval states
• Real-time endpoints, endpoint configurations, and instance selection
• Serverless inference and asynchronous inference use cases
• Batch Transform for offline prediction workloads
• Inference payload design, response handling, and endpoint testing
• Amazon CloudWatch logs, metrics, alarms, and endpoint cost controls

Workshop: Participants register their selected model, deploy it to a real-time endpoint, test prediction payloads, and write a deployment decision note.

Day 5: Monitoring, governance, and implementation planning

• SageMaker Model Monitor baselines and data-quality monitoring
• Data drift, model drift, and prediction-quality review triggers
• Model lineage, reproducibility, and audit evidence
• Secure data handling, encryption, network isolation, and IAM review
• MLOps workflow roles and approval gates
• Compute sizing, endpoint utilisation, and SageMaker cost management
• Architecture patterns for moving a proof of concept into production

Workshop: Participants present a SageMaker implementation blueprint containing an architecture diagram, operating controls, monitoring plan, ownership model, and 90-day action plan.

Tools & standards covered

Amazon SageMaker Studio, SageMaker Python SDK, Amazon S3, Amazon CloudWatch

A typical training day

Live online deliveries follow the same structure in the East Africa Time zone, with shorter screen blocks and longer breaks.

What the fee includes

How you can take this course

Certification

Participants who complete the full five days receive the Skillset Development Certificate of Completion, stating the course title, course code, dates and delivery format — suitable for professional-development records and employer reimbursement.

Frequently asked questions