AWS MLA-C01 vs MLA-C02: What Changed & Bedrock

The AWS Certified AI Practitioner exam was updated from MLA-C01 to MLA-C02, and the biggest change is a much stronger focus on Amazon Bedrock and generative AI fundamentals. In short, MLA-C02 removes most of the older machine learning (ML) lifecycle content and replaces it with topics like foundation models, prompt engineering, and Bedrock services. This article explains exactly what changed, what the Bedrock emphasis means for your study plan, and how to prepare effectively.

Why AWS Updated the Exam

AWS released MLA-C01 in late 2023 as an entry-level certification for people who work with AI and ML but are not necessarily data scientists. The exam covered a broad range of ML concepts, including data preparation, model training, and evaluation. However, by 2024, the AI landscape shifted dramatically toward generative AI and large language models (LLMs). AWS responded by updating the exam to reflect what practitioners actually do today: using managed services like Amazon Bedrock, rather than building models from scratch.

The MLA-C02 exam guide explicitly states that it targets “individuals in AI and ML roles who are responsible for using, integrating, and operationalizing generative AI applications.” This is a clear departure from MLA-C01, which was more of a general AI/ML overview. The update also aligns with AWS’s push to make Bedrock its primary platform for generative AI workloads.

Key Changes Between MLA-C01 and MLA-C02

Here are the most important differences you need to know:

• More Bedrock, less traditional ML: MLA-C02 heavily emphasizes Amazon Bedrock, including foundation models, agents, knowledge bases, and guardrails. MLA-C01 only touched on these topics lightly.
• Removed ML lifecycle details: Topics like data preparation, feature engineering, and model evaluation are largely gone. MLA-C02 assumes you understand basic ML concepts but focuses on applying them through AWS services.
• New emphasis on prompt engineering: MLA-C02 includes specific questions about prompt design, token limits, and how to improve model responses using techniques like few-shot prompting.
• Updated domain weights: The exam domains are now: Domain 1 – AI and ML Fundamentals (20%), Domain 2 – Foundations of Generative AI (24%), Domain 3 – Applications of Foundation Models (34%), Domain 4 – Responsible AI (12%), and Domain 5 – AI Security and Operations (10%). Compare this to MLA-C01, which had a heavier focus on ML fundamentals (30%) and less on generative AI.
• Bedrock-specific services: MLA-C02 expects you to know about Bedrock Agents, Knowledge Bases, Guardrails, and the Model Invocation API. MLA-C01 only mentioned Bedrock in passing.

What “Bedrock In” Means for the Exam

You might have seen the phrase “Bedrock in” in study guides or forums. It simply means that Amazon Bedrock is now a core part of the exam. In MLA-C02, you will be asked to:

• Identify which Bedrock feature to use for a given scenario (e.g., Knowledge Bases for RAG, Guardrails for safety).
• Understand how to invoke foundation models via the Bedrock API.
• Explain the difference between managed APIs and self-hosted models.
• Know the basic components of an agent (instruction, action groups, knowledge bases).

For example, a question might describe a company that wants to build a chatbot that can answer questions from its own documents. The correct answer would be to use Amazon Bedrock with a Knowledge Base, not to train a custom model. This is the kind of practical, service-oriented question that defines MLA-C02.

How to Prepare for MLA-C02

If you already studied for MLA-C01, you will need to adjust your approach. Here is a step-by-step plan:

1. Review the official exam guide – Download the MLA-C02 guide from AWS and note the domain weights. Spend more time on Domain 3 (Applications of Foundation Models) since it is the largest.
2. Learn Bedrock fundamentals – Understand what Bedrock is, how it provides access to foundation models (via API), and the key features: Agents, Knowledge Bases, Guardrails, and Provisioned Throughput.
3. Practice prompt engineering – Study techniques like zero-shot, few-shot, chain-of-thought, and how to structure prompts for different tasks (summarization, classification, extraction).
4. Do hands-on labs – If you have AWS access, try creating a simple Bedrock agent or a Knowledge Base. This will help you remember the concepts.
5. Take practice exams – Use official AWS practice questions or reputable third-party tests. Focus on scenario-based questions that ask “which service” or “which feature.”

Common Mistakes and Misconceptions

One common mistake is assuming MLA-C02 is just a renamed version of MLA-C01. It is not – the content is significantly different, and you will fail if you only study old materials. Another misconception is that you need to know how to train models. You do not. The exam is about using AWS services, not building algorithms.

Another pitfall is ignoring the Responsible AI domain (12%). Questions about bias, fairness, and model explainability are included, and they often reference Bedrock Guardrails or SageMaker Clarify. Make sure you understand the difference between these two services.

Also, do not confuse Bedrock with SageMaker. Bedrock is a managed service for foundation models; SageMaker is a broader ML platform for building and training custom models. The exam will test your ability to choose the right service for the job.

Costs and Time to Prepare

The exam fee is $100 USD, same as MLA-C01. You have 65 minutes to answer 50 questions, and the passing score is 700 out of 1000. Most candidates spend 4-6 weeks preparing, depending on their background. If you are new to AWS, you might need more time to learn basic IAM, S3, and Lambda, which are prerequisites for understanding Bedrock.

There is no official AWS training course specifically for MLA-C02 yet, but the free “AWS Skill Builder” has an exam prep plan that includes digital courses and practice questions. You can also use the free tier to experiment with Bedrock, though some features may incur costs.

Frequently Asked Questions

Is MLA-C02 harder than MLA-C01?

It is not necessarily harder, but it is more focused. If you are comfortable with generative AI concepts and AWS services, you will find it easier than the broad ML topics in MLA-C01. If you are new to AI, you may need extra time to learn the basics.

Do I need to know Python or coding for MLA-C02?

No. The exam is multiple-choice and does not require programming. However, you should understand what APIs are and how to call them, since Bedrock is accessed via API. You might see code snippets in questions, but they are usually simple and you only need to understand the intent, not write code.

Can I skip MLA-C01 and take MLA-C02 directly?

Yes, MLA-C02 is a standalone certification. There is no prerequisite. If you are new to AI, start with the fundamentals, but you do not need to pass MLA-C01 first.

Will MLA-C01 still be valid if I already passed it?

Yes, if you passed MLA-C01, your certification is valid for three years from the date you passed. You do not need to retake the exam until it expires. However, if you want to stay current, you can take MLA-C02 as a new certification.

What is the difference between Bedrock and SageMaker on the exam?

Bedrock is for using pre-trained foundation models via API. SageMaker is for building, training, and deploying custom ML models. On the exam, if the scenario involves a foundation model (like Claude or Llama) and you need to integrate it quickly, choose Bedrock. If the scenario involves training a custom model on your own data, choose SageMaker.

Are there any labs or hands-on components in the exam?

No, the exam is entirely multiple-choice with some multiple-response questions. There is no practical lab section. However, hands-on experience with Bedrock will help you answer scenario-based questions more confidently.

Related

• Galaxy S6 Edge Home Button Not Working? Fix It
• Hydration During Exercise: Water, Electrolytes, Timing
• Starlink Speed Boost: Reduce Latency & Improve