AWS exam changes / Machine Learning Engineer – Associate

MLA-C01 vs MLA-C02: what changes in the AWS Machine Learning Engineer – Associate exam

MLA-C02 replaces MLA-C01. It is in beta now, and the standard exam starts in all languages on January 14, 2027.

Last updated November 12, 2026.

Key dates

What changes in MLA-C02

MLA-C02 keeps the same four domains, the same 12 task areas and the same passing score, but every domain now covers generative AI as well as traditional machine learning. AWS’s comparison page lists about 45 new skills, almost all about foundation models, retrieval-augmented generation (RAG), vector stores and AI agents, and a handful of removed ones.

Beta figures are from AWS’s announcement. Check theofficial certification pagefor the standard exam’s length and price before you book.

Domains and weightings

Every domain name now says “ML and AI”. Model development drops from 26% to 24% and becomes “ML Model and Foundation Model (FM) Development”; deployment and orchestration rises from 22% to 24%. Data preparation stays the largest domain at 28%.

New skills, domain by domain

AWS’s comparison maps every MLA-C01 task to an MLA-C02 task with a new title, then lists what was added. In short:

• Data preparation (1.1 to 1.3): vector databases (OpenSearch Service, RDS with pgvector, S3), multimodal data, embedding models, chunking and metadata for RAG, masking and redacting data, preparing data for fine-tuning, continued pre-training and distillation, and validating prompt-response training data.
• Model and FM development (2.1 to 2.3): choosing Bedrock foundation models, fine-tuning strategies, RAG architecture patterns, trading off performance, latency and cost, prompt engineering, tuning retrieval and embeddings, human-in-the-loop evaluation, NLP metrics (BLEU, ROUGE, BERTScore), LLM-as-a-judge, and monitoring RAG retrieval accuracy.
• Deployment and orchestration (3.1 to 3.3): FM deployment and hosting options, importing models built outside AWS (Bedrock Custom Model Import), deploying agents and their communication protocols, Bedrock knowledge bases, retrieval pipelines, agent state, GPU scaling, Bedrock Prompt Management, agent and prompt testing in pipelines, and refreshing knowledge bases automatically.
• Operations, monitoring and security (4.1 to 4.3): monitoring agents (tool failures, truncated streaming), Bedrock evaluations, FM inference and token costs, embedding and vector storage costs, scanning pipelines and images (CodeGuru, Inspector), Bedrock API keys versus IAM credentials, and Bedrock Guardrails for responsible AI.

Removed from the task statements

• Configuring data to load into training resources such as Amazon EFS and Amazon FSx (1.3).
• Fine-tuning pre-trained models with custom datasets in Bedrock or SageMaker JumpStart, and reducing model size by pruning or compression (2.2). Fine-tuning itself is still in, under the new FM skills.
• Optimizing models for edge devices with SageMaker Neo (3.1).
• Bring your own container (BYOC) with SageMaker (3.2).
• Monitoring infrastructure with EventBridge events, and troubleshooting capacity with provisioned concurrency, service quotas and auto scaling (4.2).

What to study differently

Our reading of AWS’s changes, not AWS’s wording.

• Learn RAG end to end: chunking, embeddings, a vector store, Bedrock Knowledge Bases, reranking, and how to measure retrieval quality. It appears in every domain.
• Know when to prompt, when to use RAG and when to fine-tune or distill, and what each costs in latency and money.
• Learn how to evaluate generative output: human review, BLEU and ROUGE, LLM-as-a-judge and Bedrock evaluations.
• Get familiar with agents: deploying them, keeping state, versioning them in pipelines and spotting tool failures.
• Security now includes Bedrock Guardrails, Bedrock API keys versus IAM credentials, and redacting sensitive data before training.
• SageMaker AI is still the core of the exam. Training, tuning, deployment options, Model Monitor and Clarify all carry over.

The MLA-C02 exam guide

65 questions (50 scored, 15 unscored), passing score 720 out of 1,000. Multiple choice and multiple response. AWS has not yet published the standard exam’s length and price.

1. Data Preparation for ML and AI (28%)

   • Collect and store data
   • Perform data transformation, feature engineering, and pre-processing
   • Validate data quality and manage bias

2. ML Model and Foundation Model (FM) Development (24%)

   • Choose appropriate modeling approaches for ML and AI solutions
   • Train, fine-tune, and customize models for ML and AI solutions
   • Analyze and evaluate the performance of ML and AI systems

3. Deployment and Orchestration of ML and AI Workflows (24%)

   • Manage deployment infrastructure for ML and AI model types
   • Provision and configure resources for ML and AI workloads based on existing architecture and requirements
   • Implement automated orchestration and continuous integration and continuous delivery (CI/CD) pipelines for MLOps and AI workloads

4. Operating, Monitoring, and Securing ML and AI Solutions (24%)

   • Monitor ML and AI model inference and performance
   • Optimize and manage ML and AI infrastructure costs and performance
   • Secure ML and AI workloads and model endpoints

Source: AWS MLA-C02 exam guide.

The MLA-C01 exam guide

65 questions (50 scored, 15 unscored) in 130 minutes, $150, passing score 720 out of 1,000. Multiple choice and multiple response.

1. Data Preparation for Machine Learning (ML) (28%)

   • Ingest and store data
   • Transform data and perform feature engineering
   • Ensure data integrity and prepare data for modeling

2. ML Model Development (26%)

   • Choose a modeling approach
   • Train and refine models
   • Analyze model performance

3. Deployment and Orchestration of ML Workflows (22%)

   • Select deployment infrastructure based on existing architecture and requirements
   • Create and script infrastructure based on existing architecture and requirements
   • Use automated orchestration tools to set up continuous integration and continuous delivery (CI/CD) pipelines

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

   • Monitor model inference
   • Monitor and optimize infrastructure and costs
   • Secure AWS resources

Source: AWS MLA-C01 exam guide.

Which version should you take?

• In English: MLA-C01 is no longer available. The last day was September 28, 2026, so you’ll take the MLA-C02 beta now, or the standard MLA-C02 from January 14, 2027.
• In Japanese, Korean or Simplified Chinese: MLA-C01 stays available until the standard MLA-C02 starts on January 14, 2027.
• Either version earns the AWS Certified Machine Learning Engineer – Associate certification, and an MLA-C01 certification stays valid until it expires.

Practicing on ExamPark

Our free Machine Learning Engineer – Associate practice exam and theMLA-C02 study guide follow the MLA-C02 exam guide, including the generative AI skills above.

Sources

• AWS: Certification updates, September 2026
• Every announced AWS exam change, with dates

Take the free practice exam