AWS replaced the English-language MLA-C01 exam with MLA-C02 on 29 September 2026. The last day to sit MLA-C01 in English was 28 September. If you are starting your preparation now and will sit the exam in English, you are preparing for MLA-C02, and this 8-week plan is built on the MLA-C02 exam guide. MLA-C01 is still offered in Japanese, Korean and Simplified Chinese until MLA-C02 reaches general availability, so the older outline is covered below as well.
The certification itself has not changed name. It is still the AWS Certified Machine Learning Engineer - Associate, and a pass on either version earns the same credential.
MLA-C01 or MLA-C02: Which Exam Are You Sitting?
| Your situation | Exam you will sit |
|---|---|
| Sitting in English, from 29 September 2026 | MLA-C02 (in beta, general availability date not yet announced) |
| Sitting in Japanese, Korean or Simplified Chinese before MLA-C02 general availability | MLA-C01 |
| Sitting in any language after MLA-C02 general availability | MLA-C02 |
AWS lists the MLA-C02 general availability registration and delivery dates as "TBD" on the certification page. Check that page before you book, because AWS betas run for a limited period and the dates will be posted there first. Our MLA-C01 to MLA-C02 change guide covers the beta rules and the full list of added and removed content.
Exam at a Glance
| Detail | MLA-C02 (standard exam) | MLA-C02 beta | MLA-C01 |
|---|---|---|---|
| Questions | 65 (50 scored, 15 unscored) | 85 | 65 (50 scored, 15 unscored) |
| Time | Not yet published | 170 minutes | 130 minutes |
| Cost | Not yet published | $75 USD | $150 USD |
| Passing score | 720 (scaled 100 to 1,000) | Not reported on the day | 720 (scaled 100 to 1,000) |
| Question types | Multiple choice, multiple response | Multiple choice, multiple response | Multiple choice, multiple response |
| Languages | Others added at general availability | English only | Japanese, Korean, Simplified Chinese |
| Validity | 3 years | 3 years | 3 years |
The MLA-C02 exam guide states that the standard exam has 50 scored and 15 unscored questions and a minimum passing score of 720. Scoring is compensatory: you need to pass the exam overall, not each domain. AWS has not yet published a duration or price for the standard MLA-C02 exam. MLA-C01 is 130 minutes and $150, which is the usual AWS associate format.
Beta candidates do not get a result on the day. AWS says beta results typically arrive within five business days of the exam, and a beta can be attempted only once.
There are no formal prerequisites. For MLA-C02, AWS describes the target candidate as having at least one year of experience using Amazon SageMaker AI, Amazon Bedrock and other AWS services for ML engineering, plus at least one year in a related role such as backend developer, DevOps developer, data engineer or data scientist. The candidate is expected to have worked with both traditional ML and generative AI.
The Four Exam Domains
The domain structure survived the update. The weights shifted slightly and every domain picked up foundation model content.
| Domain | MLA-C02 | MLA-C01 |
|---|---|---|
| 1 | Data Preparation for ML and AI (28%) | Data Preparation for Machine Learning (28%) |
| 2 | ML Model and Foundation Model (FM) Development (24%) | ML Model Development (26%) |
| 3 | Deployment and Orchestration of ML and AI Workflows (24%) | Deployment and Orchestration of ML Workflows (22%) |
| 4 | Operating, Monitoring, and Securing ML and AI Solutions (24%) | ML Solution Monitoring, Maintenance, and Security (24%) |
Domain 1: Data Preparation for ML and AI (28%)
Still the largest domain. You ingest, transform, validate and prepare data. The traditional material remains: feature engineering, cleaning, data formats, and services such as AWS Glue, Glue DataBrew, Amazon S3 and SageMaker AI. MLA-C02 adds vector databases (OpenSearch Service, Amazon RDS with pgvector), embedding models, document preparation for Retrieval Augmented Generation such as chunking and metadata extraction, masking and redacting data, and preparing data for foundation model fine-tuning.
Domain 2: ML Model and Foundation Model Development (24%)
Choosing a modelling approach, training, tuning hyperparameters, evaluating performance and managing versions. MLA-C02 adds selecting foundation models from Amazon Bedrock, fine-tuning strategies, RAG architecture patterns, prompt engineering as a customisation technique, and evaluation metrics for generated text such as BLEU, ROUGE and BERTScore.
Domain 3: Deployment and Orchestration of ML and AI Workflows (24%)
Deployment infrastructure, endpoints, compute, auto scaling and CI/CD pipelines. This domain grew from 22% to 24%. The additions are foundation model deployment options, Amazon Bedrock Custom Model Import, Bedrock knowledge bases, deploying and configuring agents, agent state management, GPU scaling, and prompt management.
Domain 4: Operating, Monitoring, and Securing ML and AI Solutions (24%)
Monitoring models, data and infrastructure, managing cost, and securing workloads. MLA-C02 adds agent performance monitoring, Amazon Bedrock evaluations, foundation model inference cost and token usage, Amazon Bedrock Guardrails, and checking pipelines for code and image vulnerabilities with Amazon CodeGuru and Amazon Inspector.
What AWS Removed
If you are working from older MLA-C01 notes, AWS lists these as removed in MLA-C02: loading training data from Amazon EFS and Amazon FSx, reducing model size by pruning and compression, optimising models for edge devices with SageMaker Neo, bring your own container with SageMaker, monitoring infrastructure with EventBridge events, and troubleshooting capacity concerns such as provisioned concurrency and service quotas. Candidates sitting MLA-C01 in another language still need them.
Exam Tip: Domains 3 and 4 are 48% of MLA-C02. If your background is data science, that is where your revision time should go, and it is where the new agent and Bedrock content is concentrated.
Who This Exam Suits
This is an engineering exam. It tests whether you can take a model, or a foundation model, and run it in production: pipelines, deployment, automation, monitoring, cost and security. It replaced the theory-heavy AWS Certified Machine Learning - Specialty (MLS-C01), whose last exam day was 31 March 2026. Our MLS-C01 retirement guide explains where that leaves Specialty holders.
If you have not used AWS AI services before, the AWS AI Practitioner is the gentler first step, and our 4-week AIF-C01 study plan covers it.
The 8-Week Study Plan
The plan assumes 8 to 10 hours a week. Stretch it to 12 weeks if you have less time. If you use SageMaker AI and Bedrock at work, six weeks is realistic. Candidates sitting MLA-C01 in another language can follow the same weeks and skip the items marked MLA-C02.
Week 1: Exam Guide and Baseline
- Read the official MLA-C02 exam guide end to end, including the in-scope services list.
- Read AWS's MLA-C01 to MLA-C02 comparison page. It shows which older material is still valid, and it is the official record of the MLA-C01 outline for candidates still sitting that version.
- Set up an AWS account with a budget alert in AWS Budgets before you start any labs.
- Refresh the basics: supervised and unsupervised learning, training versus inference, common metrics, and what a foundation model is.
Week 2: Data Preparation (Domain 1)
- Ingest and transform data with AWS Glue, Glue DataBrew and Amazon S3. Use Glue Data Quality for validation.
- Practise feature engineering: encoding, scaling, imputing missing values, handling imbalanced data.
- Lab: load a public dataset into S3, clean it, and store features in SageMaker Feature Store.
- MLA-C02: create embeddings with a Bedrock embedding model and load them into a vector store. Try two chunking strategies on the same documents and compare retrieval results.
Week 3: Model Development, Traditional ML (Domain 2)
- Learn the SageMaker AI built-in algorithms and when each fits.
- Run training jobs and automatic model tuning.
- Read confusion matrices and choose between precision, recall and F1 for a given business problem.
- Lab: train a model and run a hyperparameter tuning job.
Week 4: Model Development, Foundation Models (Domain 2)
- Use SageMaker Clarify for bias detection and explainability, and register model versions in the Model Registry.
- MLA-C02: compare foundation models in Amazon Bedrock for one task on quality, latency and cost.
- MLA-C02: learn when to use prompt engineering, RAG or fine-tuning, and what each costs to run.
- MLA-C02: learn what BLEU, ROUGE and BERTScore measure, and where human evaluation is needed.
Week 5: Deployment (Domain 3)
- Know the SageMaker AI inference options: real-time, serverless, asynchronous and batch transform.
- Study auto scaling, multi-model endpoints and the cost trade-offs between them.
- Lab: deploy a trained model to a real-time endpoint and invoke it.
- MLA-C02: build a Bedrock knowledge base over your Week 2 documents, then deploy a simple agent that uses it.
Week 6: Orchestration and CI/CD (Domain 3)
- Orchestrate a workflow with SageMaker Pipelines and AWS Step Functions.
- Automate build and deployment with CodePipeline and CodeBuild, and define resources with CloudFormation or the AWS CDK.
- Lab: a pipeline that retrains and redeploys a model when new data lands in S3.
- MLA-C02: version prompts with Bedrock Prompt Management and add a prompt test to the pipeline.
Week 7: Monitoring, Cost and Security (Domain 4)
- Detect data drift and model drift with SageMaker Model Monitor. Use CloudWatch for metrics, logs and alarms.
- Track spend with Cost Explorer and AWS Budgets.
- Secure workloads: IAM roles, encryption with AWS KMS, Secrets Manager and VPC configuration.
- MLA-C02: configure Amazon Bedrock Guardrails, run a Bedrock evaluation, and work out the token cost of one of your Week 4 experiments.
Week 8: Timed Practice and Review
- Sit full-length timed practice exams on separate days. For the beta, practise at the beta length: 85 questions in 170 minutes.
- Review every wrong answer until you can explain why the right option is right.
- Reread the task statements in the exam guide and mark any you could not explain to a colleague.
- Spend the last two days on your weakest domain. Do not start new topics.
Mistakes That Cost Marks
- Studying from MLA-C01 or MLS-C01 material without checking it. MLS-C01 covered different objectives. MLA-C01 material is mostly still valid for MLA-C02, but it has no agents, RAG, knowledge bases or guardrails in it.
- Reading without building. Scenario questions ask which service or configuration solves a stated problem. That is much easier to answer if you have deployed an endpoint and built a knowledge base yourself.
- Neglecting deployment and operations. Domains 3 and 4 are nearly half the exam.
- Treating the beta like a standard sitting. It is longer, the result is delayed, and you get one attempt.
Where CertCrush Fits
CertCrush does not have an MLA-C02 course yet. The closest course is AWS Certified AI Practitioner (AIF-C01), which covers foundation models, Amazon Bedrock, prompt engineering and responsible AI at foundational level. It is useful groundwork for the generative AI content added in MLA-C02, and it will not prepare you for the engineering depth of the associate exam on its own.
Frequently Asked Questions
Can I still take MLA-C01?
Not in English. The last English sitting was 28 September 2026. MLA-C01 remains available in Japanese, Korean and Simplified Chinese until MLA-C02 reaches general availability.
When does MLA-C02 reach general availability?
AWS has not announced a date. The certification page lists general availability registration and delivery as "TBD".
Is my MLA-C01 certification still valid?
Yes. The certification is valid for three years from the date you earned it, and the credential name is the same for both exam versions.
Is MLA-C02 harder than MLA-C01?
AWS publishes no pass rates, so nobody can say from data. The scope is wider: MLA-C02 adds foundation models, RAG, agents and guardrails to every domain while keeping most of the traditional ML content.
Sources
- AWS Certified Machine Learning Engineer - Associate, AWS certification page
- MLA-C02 exam guide, AWS documentation
- Comparison of MLA-C01 and MLA-C02, AWS documentation
- AWS Certified Machine Learning - Specialty, AWS certification page
