AWS Certified Machine Learning - Specialty (MLS-C01) - RETIRED¶

Important Notice¶

The AWS Certified Machine Learning - Specialty (MLS-C01) certification was retired on April 15, 2025.

This certification is no longer available for new candidates. If you hold this certification, your credential will remain valid for three years from the date you earned it.

Replacement Certification¶

AWS has launched the AWS Certified Machine Learning Engineer - Associate (MLA-C01) as the modern replacement.

📖 AWS Machine Learning Engineer - Associate - Official certification page

Key Differences¶

Level: Associate (vs. Specialty for MLS-C01)
Focus: SageMaker-centric ML engineering and MLOps, with strong coverage of Bedrock and generative AI
Modern Services: Updated to include current SageMaker capabilities, Bedrock, and contemporary MLOps tooling
Career Alignment: Better aligned with today's ML engineer / MLOps roles

For Current Credential Holders¶

If you currently hold the MLS-C01 certification: - Your certification remains valid for 3 years from the date earned - Maintain your credential by earning other AWS certifications - Consider AWS Certified Machine Learning Engineer - Associate (MLA-C01) to stay current

The historical study material below is preserved for credential holders and as a reference. For new candidates, study the MLA-C01 fact sheet instead.

Historical Exam Overview (MLS-C01)¶

The AWS Certified Machine Learning - Specialty (MLS-C01) exam validated expertise in building, training, tuning, and deploying machine learning models using AWS Cloud. This certification demonstrated the ability to design, implement, deploy, and maintain machine learning solutions for given business problems.

Exam Details¶

Exam Code: MLS-C01
Duration: 180 minutes (3 hours)
Format: Multiple choice and multiple response
Number of Questions: 65 scored questions
Passing Score: 750/1000
Cost: $300 USD
Language: Available in multiple languages
Delivery: Testing center or online proctoring
Validity: 3 years
Prerequisites: 1-2 years hands-on experience developing, architecting, or running ML workloads on AWS

Exam Domains¶

Domain 1: Data Engineering (20% of scored content)¶

Create data repositories for machine learning
Identify and implement a data ingestion solution
Identify and implement a data transformation solution

Key Focus Areas:¶

Data lake and data warehouse architectures
ETL/ELT pipelines with AWS Glue and other services
Data formats and storage optimization
Real-time and batch data processing
Data versioning and lineage tracking
Feature stores and data preparation

Domain 2: Exploratory Data Analysis (24% of scored content)¶

Sanitize and prepare data for modeling
Perform feature engineering
Analyze and visualize data for machine learning

Key Focus Areas:¶

Data cleaning and preprocessing techniques
Feature engineering and selection strategies
Statistical analysis and hypothesis testing
Data visualization with Amazon QuickSight
Outlier detection and handling
Data bias identification and mitigation

Domain 3: Modeling (36% of scored content)¶

Frame business problems as machine learning problems
Select the appropriate model(s) for a given machine learning problem
Train machine learning models
Perform hyperparameter optimization
Evaluate machine learning models

Key Focus Areas:¶

ML problem formulation and approach selection
Algorithm selection for different problem types
Amazon SageMaker training jobs and built-in algorithms
Custom model development and containerization
Hyperparameter tuning strategies and automation
Model evaluation metrics and validation techniques

Domain 4: Machine Learning Implementation and Operations (20% of scored content)¶

Build machine learning solutions for performance, availability, scalability, resiliency, and fault tolerance
Recommend and implement the appropriate machine learning services and features for a given problem
Apply basic AWS security practices to machine learning solutions
Deploy and operationalize machine learning solutions

Key Focus Areas:¶

Model deployment strategies and endpoints
Auto-scaling and performance optimization
A/B testing and canary deployments
Model monitoring and drift detection
CI/CD for machine learning pipelines
Security and compliance in ML workflows

Core AWS Machine Learning Services¶

Amazon SageMaker¶

SageMaker Studio¶

Integrated Development Environment: Web-based IDE for ML development
Jupyter Notebooks: Managed Jupyter notebook instances
Experiment Management: Track and compare ML experiments
Model Registry: Centralized model repository with versioning
Feature Store: Centralized repository for ML features
Data Wrangler: Visual data preparation and feature engineering

SageMaker Training¶

Built-in Algorithms: Pre-built algorithms for common ML tasks
Bring Your Own Container: Custom algorithm containerization
Distributed Training: Multi-instance and multi-GPU training
Spot Training: Cost-effective training with EC2 Spot instances
Automatic Model Tuning: Hyperparameter optimization service
Debugger: Real-time training job monitoring and debugging

SageMaker Inference¶

Real-time Endpoints: Low-latency model serving
Batch Transform: Batch inference for large datasets
Multi-Model Endpoints: Host multiple models on single endpoint
Auto Scaling: Automatic endpoint scaling based on traffic
Inference Pipelines: Chain multiple containers for preprocessing and inference
Edge Deployment: Deploy models to edge devices with SageMaker Neo

SageMaker Pipelines¶

ML Workflows: Define and orchestrate ML workflows
Pipeline Steps: Data processing, training, and deployment steps
Conditional Execution: Dynamic workflow execution based on conditions
Caching: Step-level caching for efficiency
Integration: Native integration with other SageMaker services
Monitoring: Pipeline execution monitoring and alerting

AI Services¶

Amazon Comprehend¶

Natural Language Processing: Text analysis and insights
Sentiment Analysis: Detect sentiment in text
Entity Recognition: Identify people, places, and things
Key Phrase Extraction: Extract important phrases
Language Detection: Identify the language of text
Custom Models: Train custom NLP models

Amazon Rekognition¶

Computer Vision: Image and video analysis
Object Detection: Identify objects and scenes
Facial Analysis: Detect and analyze faces
Text Detection: Extract text from images
Content Moderation: Detect inappropriate content
Custom Labels: Train custom image classification models

Amazon Textract¶

Document Analysis: Extract text and data from documents
Form Processing: Extract key-value pairs from forms
Table Extraction: Extract structured data from tables
Handwriting Recognition: Process handwritten text
Async Processing: Handle large document processing jobs
Output Formats: Multiple output formats for downstream processing

Amazon Transcribe¶

Speech-to-Text: Convert audio to text
Real-time Streaming: Live audio transcription
Speaker Identification: Identify different speakers
Custom Vocabulary: Improve accuracy with domain-specific terms
Language Support: Multiple language support
Medical Transcription: Specialized medical vocabulary and formatting

Amazon Translate¶

Neural Machine Translation: High-quality text translation
Real-time Translation: Low-latency translation service
Batch Translation: Process large volumes of text
Custom Terminology: Domain-specific translation customization
Language Detection: Automatic source language detection
Formality Control: Adjust translation formality level

Amazon Polly¶

Text-to-Speech: Convert text to lifelike speech
Neural Voices: High-quality neural text-to-speech
SSML Support: Speech Synthesis Markup Language
Lexicons: Custom pronunciation controls
Speech Marks: Synchronize speech with visual content
Multiple Output Formats: Various audio format support

Data Processing and Analytics¶

AWS Glue¶

ETL Service: Managed extract, transform, and load service
Data Catalog: Centralized metadata repository
Crawlers: Automatic schema discovery and cataloging
Job Scheduling: Automated ETL job execution
DataBrew: Visual data preparation service
Schema Registry: Manage and evolve data schemas

Amazon Kinesis¶

Data Streaming: Real-time data ingestion and processing
Kinesis Data Streams: Scalable real-time data streaming
Kinesis Data Firehose: Load streaming data into data stores
Kinesis Analytics: Real-time analytics on streaming data
Video Streams: Capture and process video streams
Integration: ML model integration for real-time inference

Amazon EMR¶

Big Data Processing: Managed cluster platform for big data frameworks
Spark and Hadoop: Support for Apache Spark, Hadoop, and other frameworks
Notebooks: Managed Jupyter notebook environments
Auto Scaling: Automatic cluster scaling based on workload
Cost Optimization: Spot instances and reserved capacity
Integration: SageMaker and other AWS service integration

Storage and Databases¶

Amazon S3¶

Data Lake Storage: Scalable object storage for ML datasets
Storage Classes: Optimize costs with different storage classes
Data Partitioning: Optimize query performance with partitioning
Event Notifications: Trigger ML workflows on data changes
Access Patterns: Optimize for different ML access patterns
Security: Encryption and access control for ML data

Amazon DynamoDB¶

NoSQL Database: Fast, scalable NoSQL database
Feature Storage: Store ML features with low-latency access
Streams: Capture data changes for real-time ML
Global Tables: Multi-region replication for global ML applications
Performance: Consistent single-digit millisecond latency
Integration: Native integration with ML services

Machine Learning Fundamentals¶

Problem Types and Algorithms¶

Supervised Learning¶

Linear and Logistic Regression: Fundamental regression and classification
Decision Trees and Random Forest: Tree-based algorithms
Support Vector Machines: Margin-based classification
Neural Networks: Deep learning approaches
Ensemble Methods: Combining multiple models
Time Series Forecasting: Temporal data prediction

Unsupervised Learning¶

Clustering: K-means, hierarchical, and density-based clustering
Dimensionality Reduction: PCA, t-SNE, and autoencoders
Anomaly Detection: Statistical and ML-based approaches
Association Rules: Market basket analysis and recommendations
Topic Modeling: Text analysis and document clustering
Generative Models: GANs and variational autoencoders

Reinforcement Learning¶

Q-Learning: Value-based reinforcement learning
Policy Gradient: Policy-based approaches
Actor-Critic: Hybrid value and policy methods
Multi-Armed Bandits: Exploration vs exploitation
Deep Reinforcement Learning: Neural network-based RL
Applications: Game playing, robotics, and optimization

Feature Engineering¶

Data Preprocessing¶

Data Cleaning: Handle missing values, outliers, and inconsistencies
Data Normalization: Scaling and standardization techniques
Categorical Encoding: One-hot encoding, label encoding, target encoding
Text Processing: Tokenization, stemming, and vectorization
Image Processing: Resizing, normalization, and augmentation
Time Series: Seasonality, trends, and lag features

Feature Selection¶

Filter Methods: Statistical tests and correlation analysis
Wrapper Methods: Forward/backward selection and recursive elimination
Embedded Methods: L1/L2 regularization and tree-based importance
Dimensionality Reduction: PCA, LDA, and feature extraction
Domain Knowledge: Expert-driven feature creation
Automated Feature Engineering: Tools and techniques for automation

Model Evaluation and Validation¶

Evaluation Metrics¶

Classification: Accuracy, precision, recall, F1-score, AUC-ROC
Regression: MAE, MSE, RMSE, R-squared, MAPE
Ranking: NDCG, MAP, MRR
Clustering: Silhouette score, Davies-Bouldin index
Time Series: MASE, sMAPE, MSIS
Business Metrics: Custom metrics aligned with business objectives

Validation Techniques¶

Cross-Validation: K-fold, stratified, and time series cross-validation
Hold-out Validation: Train/validation/test splits
Bootstrap Sampling: Resampling techniques for uncertainty estimation
Nested Cross-Validation: Hyperparameter tuning with proper validation
Temporal Validation: Time-aware validation for temporal data
A/B Testing: Online evaluation and experimentation

Hyperparameter Optimization¶

Search Strategies¶

Grid Search: Exhaustive search over parameter grid
Random Search: Random sampling of parameter space
Bayesian Optimization: Efficient search using Gaussian processes
Evolutionary Algorithms: Genetic algorithms and particle swarm optimization
Multi-Fidelity: Early stopping and successive halving
Neural Architecture Search: Automated neural network design

SageMaker Automatic Model Tuning¶

Hyperparameter Tuning Jobs: Automated hyperparameter optimization
Search Strategies: Bayesian, random, and grid search
Early Stopping: Automatic stopping of unpromising training jobs
Warm Start: Initialize tuning with previous results
Multi-Objective: Optimize for multiple metrics simultaneously
Integration: Seamless integration with SageMaker training

MLOps and Production Deployment¶

Model Deployment Strategies¶

Real-time Inference¶

SageMaker Endpoints: Managed model hosting with auto-scaling
Multi-Model Endpoints: Cost-effective hosting for multiple models
Inference Pipelines: Chaining preprocessing and inference
Container Optimization: Performance tuning for inference containers
Load Testing: Ensuring performance under production load
Monitoring: Real-time monitoring of inference performance

Batch Inference¶

Batch Transform Jobs: Large-scale batch inference
Distributed Processing: Parallel processing for large datasets
Cost Optimization: Using Spot instances for batch jobs
Scheduling: Automated batch job scheduling and orchestration
Output Management: Efficient handling of batch inference results
Error Handling: Robust error handling and retry mechanisms

Edge Deployment¶

SageMaker Neo: Model optimization for edge devices
IoT Greengrass: Deploy ML models to IoT devices
Model Compression: Techniques to reduce model size
Quantization: Reduce model precision for edge deployment
Hardware Optimization: Leverage specialized hardware accelerators
Offline Inference: Deploy models for offline operation

Model Monitoring and Management¶

Model Performance Monitoring¶

Data Drift Detection: Monitor changes in input data distribution
Model Drift Detection: Monitor degradation in model performance
Concept Drift: Detect changes in the relationship between features and target
Feature Attribution: Understanding model predictions and decisions
Bias Detection: Monitor for bias in model predictions
Explainability: Techniques for model interpretation and explanation

Model Versioning and Registry¶

Model Registry: Centralized repository for model artifacts
Version Control: Track model versions and lineage
Model Approval: Automated and manual model approval workflows
A/B Testing: Compare model versions in production
Rollback Mechanisms: Quick rollback to previous model versions
Metadata Management: Track model metrics, hyperparameters, and performance

CI/CD for Machine Learning¶

MLOps Pipelines¶

Code Integration: Version control for ML code and experiments
Automated Testing: Unit tests, integration tests, and model validation
Pipeline Orchestration: Automated ML workflow execution
Artifact Management: Track datasets, models, and experiment results
Deployment Automation: Automated model deployment and promotion
Monitoring Integration: Automated monitoring and alerting setup

Infrastructure as Code¶

CloudFormation: Infrastructure automation for ML workloads
CDK: Programmatic infrastructure definition and deployment
Terraform: Multi-cloud infrastructure automation
Container Orchestration: Kubernetes and ECS for ML workloads
Serverless ML: Lambda and serverless architectures for ML
Cost Management: Automated cost optimization and resource management

Study Strategy¶

Prerequisites¶

Statistics and Mathematics: Strong foundation in statistics, linear algebra, and calculus
Programming Skills: Python/R programming with ML libraries (scikit-learn, pandas, numpy)
AWS Foundation: AWS Associate-level knowledge and hands-on experience
ML Experience: 1-2 years of hands-on ML project experience
Data Engineering: Understanding of data pipelines and processing

Phase 1: AWS ML Services (4-5 weeks)¶

SageMaker Deep Dive
Studio environment and notebook management
Training jobs and built-in algorithms
Hyperparameter tuning and experiments
Model deployment and inference
AI Services
Comprehend for NLP tasks
Rekognition for computer vision
Textract for document processing
Transcribe and Translate for multimedia

Phase 2: Data Engineering and Processing (3-4 weeks)¶

Data Pipeline Design
AWS Glue for ETL operations
Kinesis for real-time processing
EMR for big data processing
S3 optimization for ML workloads
Feature Engineering
Data preparation techniques
Feature selection and extraction
SageMaker Data Wrangler
Feature store implementation

Phase 3: Advanced ML Concepts (4-5 weeks)¶

Algorithm Selection and Implementation
Supervised learning algorithms
Unsupervised learning techniques
Deep learning with SageMaker
Custom algorithm development
Model Optimization
Hyperparameter tuning strategies
Model evaluation and validation
Performance optimization
Bias detection and mitigation

Phase 4: MLOps and Production (2-3 weeks)¶

Deployment and Monitoring
Production deployment strategies
Model monitoring and drift detection
A/B testing and experimentation
Performance optimization
Automation and Orchestration
ML pipeline automation
CI/CD for ML workflows
Infrastructure as code
Cost optimization strategies

Hands-on Lab Recommendations¶

Essential ML Labs¶

End-to-End ML Pipeline
Data ingestion and preprocessing
Feature engineering and selection
Model training and hyperparameter tuning
Deployment and monitoring
Computer Vision Project
Image classification with SageMaker
Custom model training with transfer learning
Real-time inference endpoint
Performance optimization and monitoring
Natural Language Processing
Text classification and sentiment analysis
Custom entity recognition
Document processing pipeline
Multi-language support implementation
Time Series Forecasting
Data preparation for time series
Feature engineering for temporal data
Model selection and evaluation
Production deployment and monitoring

Exam Preparation Tips¶

Study Resources¶

AWS ML Whitepapers: Machine learning best practices and architectures
SageMaker Documentation: Comprehensive service documentation
Research Papers: Stay current with ML research and techniques
Hands-on Tutorials: AWS ML tutorials and workshops
Re:Invent Sessions: Latest ML features and case studies

Practical Experience¶

Real Projects: Work on end-to-end ML projects
Kaggle Competitions: Practice with diverse datasets and problems
Open Source Contributions: Contribute to ML open source projects
Industry Use Cases: Study ML applications in different industries
Performance Optimization: Experience with production ML systems

Exam Strategy¶

Problem Analysis: Identify ML problem type and appropriate approach
Service Selection: Choose the right AWS service for each use case
Best Practices: Apply ML and AWS best practices
Cost Optimization: Consider cost implications of different approaches
Scalability: Design for production scale and performance

Common Exam Scenarios¶

Business Problem Formulation¶

Converting business requirements to ML problems
Selecting appropriate algorithms and approaches
Defining success metrics and evaluation criteria
Considering data availability and quality constraints

Data Engineering and Preparation¶

Designing data pipelines for ML workloads
Implementing feature engineering strategies
Handling missing data and outliers
Optimizing data storage and access patterns

Model Development and Training¶

Selecting appropriate algorithms for different problem types
Implementing hyperparameter optimization strategies
Handling imbalanced datasets and bias
Evaluating model performance and generalization

Production Deployment and Operations¶

Designing scalable inference architectures
Implementing model monitoring and drift detection
Setting up A/B testing and experimentation
Automating ML workflows and operations

Next Steps After Certification¶

Career Advancement¶

ML engineer and data scientist roles
ML architect and consultant positions
Research scientist and applied researcher roles
ML platform and infrastructure engineering

Continuous Learning¶

Advanced ML techniques and research
Other AWS specialty certifications
Industry-specific ML applications
Emerging technologies (MLOps, AutoML, etc.)

Professional Development¶

ML community participation and contribution
Conference speaking and content creation
Open-source ML project leadership
Mentoring and knowledge sharing in ML teams