AWS Certified Machine Learning - Specialty Legacy Study Guide 2026

The AWS Certified Machine Learning - Specialty (MLS-C01) is now a retired certification. AWS ended the exam on March 31, 2026. That means you should not use this guide as your primary plan for a new active AWS certification path unless you are comparing legacy scope, supporting an existing certification, or mapping old MLS-C01 study material to the newer AWS machine learning track.

Even so, MLS-C01 is still worth understanding because it captures how AWS historically tested machine learning depth: data engineering, exploratory analysis, modeling, and ML implementation/operations. Compared with the newer ML Engineer Associate exam, MLS-C01 put more direct weight on classic model selection, exploratory analysis, and data-science-style workflow judgment.

This guide is therefore written as a legacy study and transition resource. It shows you what AWS officially tested on MLS-C01, where the old official documentation still lives, and how to think about the exam's scope relative to current AWS ML credentials.

Exam At a Glance

• Official certification page with retirement notice: AWS Certified Machine Learning - Specialty
• Official legacy exam guide: AWS Certified Machine Learning - Specialty exam guide
• Official legacy in-scope services reference: MLS-C01 in-scope AWS services
• Current active successor path: AWS Certified Machine Learning Engineer - Associate

Official Exam Domains

1. Data Engineering (20%)
2. Exploratory Data Analysis (24%)
3. Modeling (36%)
4. Machine Learning Implementation and Operations (20%)

The weighting explains the personality of MLS-C01. Modeling was by far the largest domain, which made the exam more data-science heavy than today's AWS ML Engineer Associate blueprint. If you are comparing the old and new exams, this is the main shift to keep in mind.

1. Data Engineering

This domain focused on building ML-ready data pipelines: where data lives, how it is ingested, and how it is transformed before modeling.

• Create the right data repositories for ML - Study storage choices across Amazon S3, databases, file storage, and training-data locality decisions. Official docs: MLS-C01 Domain 1 objectives, Amazon S3 User Guide.
• Implement batch and streaming ingestion pipelines - The official tasks include Kinesis, Data Firehose, EMR, Glue, Flink-style streaming, and job scheduling choices. Official docs: Task 1.2: Identify and implement a data ingestion solution, Amazon Kinesis Data Streams, What is AWS Glue?.
• Transform data for downstream ML workflows - Know ETL, Spark-style processing, MapReduce concepts, feature-ready transformations, and workload placement across Glue, EMR, and Batch. Official docs: Task 1.3: Identify and implement a data transformation solution, What is Amazon EMR?.
• Data engineering was foundational, not optional - MLS-C01 expected you to understand how the pipeline shape affects later model quality and training cost, not just how to store files.
• This is one of the biggest links to the modern AWS ML path - Much of the old data engineering material still transfers cleanly into current ML engineering work.

Exam tip: If an MLS-style scenario starts with poor data placement, bad ingestion flow, or the wrong transformation approach, fix the pipeline before you think about algorithms.

2. Exploratory Data Analysis

This domain is one of the clearest differences between MLS-C01 and the newer associate blueprint. AWS explicitly tested data cleaning, feature engineering, and exploratory reasoning before modeling.

• Sanitize and prepare datasets for modeling - Study missing values, corruption, normalization, scaling, augmentation, labeled data sufficiency, and mitigation strategies. Official docs: MLS-C01 Domain 2 objectives.
• Perform feature engineering across multiple data types - The official tasks cover text, speech, image, and public datasets plus encoding, outlier handling, tokenization, dimensionality reduction, and synthetic features. Official docs: Task 2.2: Perform feature engineering.
• Analyze and visualize datasets for ML fit - Know scatter plots, time series, histograms, box plots, cluster analysis, correlation, and descriptive statistics. Official docs: Task 2.3: Analyze and visualize data for ML.
• EDA was a decision-making domain - AWS was testing whether you could decide if the data was even suitable for the intended ML problem, not just clean it mechanically.
• This is one place where MLS-C01 felt closer to data science interviews - If you are transitioning to MLA-C01, expect less explicit EDA weighting and more production-oriented framing.

Exam tip: On legacy MLS-style questions, the correct answer often comes from the data distribution or feature quality rather than from the model family.

3. Modeling

This was the core of MLS-C01. The exam explicitly centered on framing the business problem, selecting model types, training, tuning, and evaluating results.

• Frame business problems as ML problems - AWS expected you to know when not to use ML and how to choose among classification, regression, forecasting, clustering, recommendation, and even foundation models in the current legacy guide version. Official docs: MLS-C01 Domain 3 objectives.
• Select the right model family - The official tasks cover XGBoost, logistic regression, k-means, linear regression, decision trees, random forests, RNNs, CNNs, ensembles, transfer learning, and LLMs. Official docs: Task 3.2: Select the appropriate model(s).
• Train and tune models effectively - Study cross-validation, compute selection, distributed versus non-distributed training, retraining patterns, optimization techniques, and hyperparameter control. Official docs: Task 3.3 and Task 3.4 objectives, Overview of machine learning with Amazon SageMaker AI.
• Evaluate models with the right metrics - The official blueprint covers AUC-ROC, accuracy, precision, recall, RMSE, F1, confusion matrices, offline and online evaluation, and overfitting versus underfitting. Official docs: Task 3.5: Evaluate ML models.
• Modeling was the biggest difference-maker on MLS-C01 - If you are using this guide to compare old and new AWS ML exams, this is where the legacy specialty stayed most academic and model-centric.

Exam tip: When an MLS-C01 scenario gives you multiple plausible model options, compare them by problem type, data characteristics, evaluation metric, and operational cost, not by popularity.

4. Machine Learning Implementation and Operations

This domain covered how ML workloads become reliable AWS systems: performance, scalability, security, and operationalization.

• Build ML solutions for scale and fault tolerance - Study CloudWatch, CloudTrail, multi-Region and multi-AZ thinking, Docker containers, Auto Scaling groups, rightsizing, and load balancing. Official docs: MLS-C01 Domain 4 objectives, What is Amazon CloudWatch?.
• Choose the right AWS ML service or feature - The old outline still expected you to choose between custom models, SageMaker built-in algorithms, and application-level AI services such as Lex, Polly, and Transcribe. Official docs: Task 4.2: Recommend and implement the appropriate ML services and features, What is Amazon SageMaker AI?.
• Apply basic AWS security to ML systems - The official tasks included IAM, S3 bucket policies, security groups, VPCs, encryption, and anonymization. Official docs: Task 4.3: Apply basic AWS security practices to ML solutions, What is IAM?.
• Deploy and operationalize models - AWS expected endpoint exposure, A/B testing, retraining pipelines, debugging, and performance monitoring. Official docs: Task 4.4: Deploy and operationalize ML solutions.
• This domain is where MLS-C01 overlaps most with the modern MLA-C01 path - If you are transitioning forward, this is one of the best sections to map directly into current study work.

Exam tip: Legacy MLS-C01 operations questions usually wanted a production-safe answer, but they still assumed more modeling fluency than the current associate exam.

How To Use This Guide In 2026

• If you already earned MLS-C01, use this guide as a scope reference for what the certification actually represented.
• If you have older MLS-C01 study material, use this guide to map it into current AWS ML concepts instead of studying the retired blueprint blindly.
• If you want an active AWS ML certification today, prioritize the AWS Machine Learning Engineer - Associate study guide and treat MLS-C01 as historical context.
• If you want a lighter AI entry point before ML engineering, use the AWS AI Practitioner study guide.

Recommended 5-Week Legacy Review Plan

Last-Mile Strategy

• Treat MLS-C01 as legacy scope documentation, not as the default AWS ML path for new candidates in 2026.
• If you are comparing MLS-C01 to MLA-C01, focus on the shift from heavier modeling and EDA weighting toward more production-oriented ML engineering.
• Use the old official domain pages to preserve accuracy when reading legacy prep material, especially for modeling and EDA terms.
• Prefer modern follow-up study on SageMaker AI operations, monitoring, security, and deployment if your goal is current job relevance.
• Do not assume that a high-quality MLS-C01 resource is current simply because it still exists online. Check the retirement status first.

If you still want legacy exam-style reinforcement, use our AWS Machine Learning Specialty practice questions. If your real goal is an active certification path in 2026, move next to our AWS Machine Learning Engineer Associate practice questions and the companion MLA-C01 study guide.

The practical value of MLS-C01 in 2026 is historical and transitional: it helps you understand the older AWS ML specialty scope, preserve context from legacy prep material, and bridge into today's active AWS machine learning certifications with clearer expectations.