MLA-C02 Study Plan: 4-Week and 8-Week

The short version

Pick the 8-week plan if you are new to SageMaker AI or Amazon Bedrock. Pick the 4-week plan only if you already use them at work and need to fill gaps. The exam weights data preparation at 28% and the other three domains at 24% each, so time is spread fairly evenly. These hours are rough, and your own pace may differ.

Before you start

  1. Read the exam guide once. AWS publishes the domains, tasks, and skills for MLA-C02. Print the skills list and mark each one green, yellow, or red. That is your map.
  2. Open an AWS account if you can. A small amount of hands-on time with SageMaker AI and Amazon Bedrock makes the service choices feel real. Watch costs: shut down endpoints and notebooks when you finish, and set a budget alert.
  3. Use current material. MLA-C02 adds retrieval augmented generation (RAG), embeddings, vector stores, Bedrock features, and agents. Older MLA-C01 notes miss those.
  4. Check the beta status. The exam was a beta as of October 18, 2026. Read the exam format page before you pick a date.

Our general advice on routines is on the how to study for IT exams page.

The 4-week plan (about 10 to 14 hours per week)

This plan assumes you already deploy or train models on AWS.

Week 1: Data preparation (28%). Storage and streaming ingestion, file formats, Glue and Spark transformations, feature engineering, SageMaker Feature Store. Add the new material: embeddings, vector databases, document chunking for RAG. Finish with data quality, bias, and class imbalance.

Week 2: Model and FM development (24%). Built-in algorithms and when each fits, hyperparameter tuning, overfitting, evaluation metrics (including precision, recall, AUC, and for text: BLEU, ROUGE), fine-tuning versus prompting versus RAG, and Bedrock model selection.

Week 3: Deployment and orchestration (24%). Endpoint types (real-time, serverless, asynchronous, batch), multi-model and multi-container hosting, deployment strategies, auto scaling metrics, VPC settings, containers, CI/CD with CodePipeline, model registries, and agents and knowledge bases.

Week 4: Operations, monitoring, and security (24%), then review. Drift detection, CloudWatch and X-Ray, cost tools, IAM least privilege, encryption, CloudTrail logging, and Bedrock Guardrails. Spend the last two days on your red and yellow items, then do a full-length practice set.

The 8-week plan (about 5 to 7 hours per week)

Week 1: Foundations. Review core ML ideas: supervised versus unsupervised learning, common algorithms, metrics, and the vocabulary of overfitting. Skim the exam guide and do your green/yellow/red map.

Week 2: Data ingestion and storage. S3, streaming with Kinesis and Flink, formats like Parquet, and choosing storage by cost and access pattern.

Week 3: Data transformation and quality. Glue, DataBrew, Data Wrangler, Feature Store, bias, imbalance, and masking sensitive data. Add embeddings and RAG document preparation.

Week 4: Training and tuning. SageMaker AI algorithms, script mode, tuning jobs, early stopping, distributed training, and debugging convergence.

Week 5: Foundation models and evaluation. Bedrock model selection, fine-tuning strategies, RAG patterns, evaluation metrics, human review, and experiment tracking.

Week 6: Deployment. Endpoints, inference options, scaling, containers, VPCs, and the automation services around them.

Week 7: Pipelines, monitoring, and cost. CI/CD for ML, model registries, retraining triggers, drift and quality monitoring, and cost tools.

Week 8: Security and review. IAM, encryption, network isolation, auditing, guardrails, then a full review and timed practice.

How to use practice exams

Practice exams measure readiness. They do not replace AWS documentation or hands-on work.

  • Take one early, cold. A baseline score shows which domain is your weakest. Do not worry if it is low.
  • Review every answer. Read why the wrong options are wrong. In this exam, the wrong options are often real services that fit a different requirement.
  • Group your misses. If three misses involve auto scaling, that is a topic, not bad luck.
  • Save a full-length set for the end. Take it in one sitting, with no notes, in the time AWS gives you. Our MLA-C02 practice tests have five sets, with the last one built as a full-length run.
  • Aim for consistent scores, not a single good score, in the range you set for yourself before you book.

Our question-writing page explains how the practice questions are made.

The last 3 days

  • Three days out: redo only the questions you missed. Skim the exam guide’s skill list once more.
  • Two days out: review service-comparison notes: which service does X and why not its neighbor.
  • The day before: light review for an hour at most. Confirm your appointment, ID, and, for online exams, your equipment and room. The exam day guide has a checklist.
  • Exam day: answer every question, because there is no penalty for guessing. Mark the long ones and return.

More guides for MLA-C02

Next step

Open the domains explained page and match each study week to a domain. Then try the sample questions to see where you stand.

Frequently asked questions

How many hours do I need?

It varies a lot. People with a year of SageMaker AI and Bedrock experience often need less than the 4-week plan above, while newcomers may need more than the 8-week plan.

Do I need hands-on labs?

They help a lot, because many questions describe real configuration choices. Even a few small projects make the services easier to tell apart.

Should I wait for the standard exam?

That is your call. The beta can be taken only once, and an AWS blog from September 2026 gives January 14, 2027 for the start of general-availability delivery, while the AWS certification page still said TBD on October 27, 2026.

Try these practice exams