MLA-C01 Study Guide: AWS Machine Learning Engineer

How to Pass the AWS Machine Learning Engineer (MLA-C01) Exam: Expert Study Guide — hero

MLA-C01 (AWS Certified Machine Learning Engineer - Associate) is a 130-minute, 65-question exam that costs USD 150 and passes at 720 out of 1,000. It ended in English on September 28, 2026. Japanese, Korean and Simplified Chinese delivery continues until January 14, 2027. AWS’s exam page lists MLA-C02 general availability as TBD; AWS’s September 2026 announcement gives January 14, 2027 for MLA-C02 general availability.

The AWS certification hub lists every AWS exam guide and practice app.

If you sit in English now, you take MLA-C02 (in beta until general availability). The four domains are similar, with generative AI added to each. Exam facts are from the AWS pages under Sources, checked October 15, 2026.

This guide covers the MLA-C01 domains, the services behind them, an 8-week plan and how to practise. Most of it also applies to MLA-C02; the section on C02 lists what changes.

What is the MLA-C01 exam format?

Exam update (October 2026): AWS ended English delivery of MLA-C01 on September 28, 2026. Japanese, Korean and Simplified Chinese delivery continues until January 14, 2027, when MLA-C01 retires; MLA-C02 general availability is announced for the same date, while the exam page lists TBD (domains 28/24/24/24%, adding generative AI topics such as Amazon Bedrock, RAG and LLM evaluation). This guide and the StudyKits MLA app follow MLA-C01; if you are booking MLA-C02, use the MLA-C02 exam guide.

MLA-C01 is an associate-level certification. AWS retired the older Machine Learning - Specialty exam (MLS-C01) on March 31, 2026. It tests your ability to build end-to-end ML solutions on AWS, from data preparation through deployment and monitoring.

The exam has 65 questions (50 scored, 15 unscored) and you get 130 minutes to complete it. You need a scaled score of 720 out of 1000 to pass. The exam costs USD 150 (JPY 20,000 in Japan).

Plan for real hands-on experience with Amazon SageMaker and related services. The scenario questions are much harder without it.

What are the four MLA-C01 domains?

Domain 1: Data Preparation for Machine Learning (28%)

This is the largest domain and the one that surprises many candidates. Nearly a third of the exam focuses on data, not models. You need to understand:

Data ingestion and transformation:

Amazon S3 as the central data lake — storage classes, lifecycle policies, partitioning strategies
AWS Glue for ETL — crawlers, jobs, data catalog, schema evolution, job bookmarks
Amazon Kinesis for streaming data — Kinesis Data Streams, Data Firehose, Data Analytics
Amazon EMR for large-scale data processing — Spark on EMR, cluster configurations
AWS Step Functions for orchestrating data pipelines

Data preparation and feature engineering:

SageMaker Data Wrangler for visual data preparation
SageMaker Feature Store — online and offline stores, feature groups, point-in-time lookups
SageMaker Processing jobs for custom data transformations
Handling imbalanced datasets — SMOTE, undersampling, oversampling, class weights
Feature encoding — one-hot encoding, label encoding, target encoding, embeddings

Data quality and governance:

AWS Glue Data Quality for automated data quality rules
Data versioning strategies
Data lineage tracking
PII detection and handling with Amazon Macie and Comprehend

Invest serious time in this domain. Many ML engineers spend most of their time in Jupyter notebooks and underestimate the data engineering questions on this exam.

Domain 2: ML Model Development (26%)

This domain tests your ability to select, train, and evaluate models.

Algorithm selection:

Know when to use SageMaker built-in algorithms: XGBoost, Linear Learner, KNN, K-Means, Random Cut Forest, DeepAR, Seq2Seq, BlazingText, Object Detection, Semantic Segmentation
Understand the tradeoffs: accuracy vs interpretability, training time vs performance
Know when to bring your own algorithm or container

Training on SageMaker:

Training jobs — instance types, distributed training (data parallelism, model parallelism)
SageMaker Training Compiler for optimizing training speed
Managed Spot Training for cost reduction
Automatic Model Tuning (hyperparameter optimization) — random search, Bayesian optimization
SageMaker Experiments for tracking training runs

Model evaluation:

Classification metrics: accuracy, precision, recall, F1, AUC-ROC, confusion matrix
Regression metrics: MSE, RMSE, MAE, R-squared
Cross-validation strategies
Bias detection with SageMaker Clarify
SageMaker Model Monitor for establishing baselines

Generative AI model development:

Fine-tuning foundation models on Amazon Bedrock
SageMaker JumpStart for deploying and fine-tuning pre-trained models
Understanding when to fine-tune vs use RAG vs prompt engineering
Model evaluation for generative AI: BLEU, ROUGE, human evaluation

Domain 3: Deployment and Orchestration of ML Workflows (22%)

This domain tests production ML skills.

Model deployment:

SageMaker real-time endpoints — instance types, auto-scaling, multi-model endpoints, multi-container endpoints
SageMaker Serverless Inference for intermittent traffic
SageMaker Batch Transform for offline predictions
SageMaker Asynchronous Inference for large payloads
Amazon Bedrock model deployment and inference
Containerizing models with Docker for SageMaker

MLOps and CI/CD:

SageMaker Pipelines for building ML workflows
SageMaker Model Registry for model versioning and approval workflows
SageMaker Projects for MLOps templates
AWS CodePipeline and CodeBuild for CI/CD
Infrastructure as Code with CloudFormation or CDK for ML infrastructure
AWS Step Functions for orchestrating end-to-end ML workflows

Deployment strategies:

Blue/green deployments with SageMaker
Canary deployments — gradual traffic shifting
Shadow testing / shadow mode
A/B testing with production variants
Rollback strategies

Domain 4: ML Solution Monitoring, Maintenance, and Security (24%)

Smaller but critical — this domain tests post-deployment skills.

Model monitoring:

SageMaker Model Monitor — data quality, model quality, bias drift, feature attribution drift
Setting up monitoring schedules and alerts
CloudWatch metrics and alarms for ML endpoints
Detecting data drift and concept drift
Automated retraining triggers

Operational monitoring:

Endpoint performance monitoring — latency, throughput, error rates
Cost optimization — right-sizing instances, auto-scaling policies, Savings Plans for SageMaker
Logging with CloudWatch Logs and CloudTrail
Troubleshooting inference errors

Model maintenance:

Retraining strategies — scheduled, triggered, continuous
Model lineage tracking
A/B testing for model updates
Feature store updates and versioning

Which AWS services should you know?

You must know these services and how they interact:

Service	Primary Use in ML
Amazon SageMaker	The core ML platform — training, deployment, monitoring
Amazon Bedrock	Managed foundation model access and fine-tuning
AWS Glue	ETL, data catalog, data quality
Amazon S3	Data storage, model artifacts, training data
Amazon EMR	Large-scale data processing with Spark
Amazon Kinesis	Real-time data streaming
AWS Step Functions	Workflow orchestration
AWS Lambda	Serverless compute for lightweight ML inference
Amazon ECR	Container registry for custom ML containers
Amazon CloudWatch	Monitoring and alerting
AWS IAM	Access control for ML resources
AWS KMS	Encryption for data and model artifacts
SageMaker Clarify	Bias detection and explainability
SageMaker Feature Store	Centralized feature management
SageMaker Pipelines	ML workflow automation
SageMaker Model Registry	Model versioning and governance

What does an 8-week study plan look like?

Weeks 1-2: Data Engineering for ML

Week 1: Study S3 data lake patterns, AWS Glue (crawlers, jobs, data catalog), and Kinesis data streaming. Do hands-on labs with Glue ETL jobs.
Week 2: Study SageMaker Data Wrangler, Feature Store, and Processing jobs. Practice feature engineering techniques. Complete 4 practice question sets on data engineering.

Weeks 3-4: Model Development

Week 3: Study SageMaker built-in algorithms (XGBoost, Linear Learner, K-Means, DeepAR, BlazingText). Understand when to use each one. Do a hands-on lab training at least two different algorithms.
Week 4: Study training job configurations, distributed training, hyperparameter tuning, and model evaluation metrics. Study SageMaker Clarify for bias detection. Complete 4 practice question sets on model development.

Weeks 5-6: Deployment and MLOps

Week 5: Study SageMaker endpoint types (real-time, serverless, batch, async). Practice deploying a model to a real-time endpoint. Study deployment strategies (blue/green, canary, shadow).
Week 6: Study SageMaker Pipelines, Model Registry, and CI/CD integration. Understand end-to-end MLOps workflows. Complete 4 practice question sets on deployment and orchestration.

Week 7: Monitoring, Maintenance, and Generative AI

Days 1-3: Study SageMaker Model Monitor (data quality, model quality, bias drift). Study CloudWatch integration and auto-scaling for endpoints.
Days 4-5: Study Bedrock fine-tuning, RAG with Bedrock Knowledge Bases, and generative AI model evaluation. Complete 3 practice question sets on monitoring and generative AI.

Week 8: Review and Exam Simulation

Days 1-2: Take a full-length 65-question practice exam under timed conditions. Identify your weakest domains.
Days 3-4: Targeted review of weak areas. Focus on services and concepts you consistently get wrong.
Day 5: Take a second full-length practice exam. Aim for 80% or higher.

How should you practise for MLA?

The MLA-C01 questions are scenario-heavy. They describe a real-world situation and ask you to choose the best solution. Reading speed and pattern recognition matter.

Build service mapping intuition. For every AWS ML service, you should instantly know its primary use case. “We need to process a large dataset” = Glue or EMR. “We need real-time predictions with auto-scaling” = SageMaker real-time endpoint. “We need to detect bias in training data” = SageMaker Clarify.

Watch for cost and operational overhead. Many questions have multiple technically correct answers, but one is more cost-effective or requires less operational overhead. AWS favors managed services over custom solutions.

Practice with explanations. Read the explanation for every question, including those you got right. It shows the reasoning AWS expects. AWS Skill Builder has official practice material for the exam.

Should you take MLA or AI Practitioner first?

If you are deciding between these two certifications, the answer depends on your role and goals. We cover this in detail in our MLA-C01 vs AI Practitioner comparison, but the short version is:

AI Practitioner (AIF-C01) is for anyone who works with AI — managers, analysts, developers, consultants
MLA-C01 is specifically for engineers who build and deploy ML systems

If you have the engineering background, MLA-C01 is the more valuable certification. But starting with AI Practitioner is a smart strategy if you are new to AWS AI services.

Where should MLA-C01 prep focus first?

The MLA-C01 is a substantial exam. It requires both theoretical knowledge and practical experience with AWS ML services. You cannot pass this exam through memorization alone — you need to understand how these services work together to solve real problems.

Start with the data engineering domain. It is the largest and the most commonly underestimated. Build hands-on experience with SageMaker by working through at least one complete ML project on AWS. And use practice questions throughout your preparation, not just at the end.

Follow the 8-week plan and practise a little every day.

What changes in MLA-C02?

MLA-C02 keeps four domains with new weights: data preparation for ML and AI (28%), ML model and foundation model development (24%), deployment and orchestration of ML and AI workflows (24%), and operating, monitoring and securing ML and AI solutions (24%).

It adds generative and agentic AI, Bedrock knowledge bases, Guardrails and Prompt Management, RAG, vector databases and LLM evaluation (BLEU, ROUGE, BERTScore, LLM-as-a-judge). It removes SageMaker Neo, bring-your-own-container and model-size reduction. The passing score stays 720 for the generally available exam. The beta is 170 minutes, 85 questions and USD 75.

The StudyKits ML Engineer app covers MLA-C01, not MLA-C02.

Frequently asked questions

Can I still take MLA-C01?

In Japanese, Korean or Simplified Chinese, until January 14, 2027. The English version ended on September 28, 2026.

What is the MLA-C01 passing score?

720 on a scale of 100-1,000.

How much does the MLA exam cost?

USD 150 for MLA-C01 and for MLA-C02 once generally available. The MLA-C02 beta is USD 75.

How to Pass the AWS Machine Learning Engineer (MLA-C01) Exam: Expert Study Guide