All about the new AWS Certified AI Practitioner exam (AIF-C01)
AWS's brand new AI certification is perfect for anyone looking to become familiar with AWS's AI and ML solutions, even if they don't actively build with them.
Sep 26, 2024 • 5 Minute Read
In the past, if you wanted to get a dedicated AI certification from AWS, you’d have to study for the infamously hard AWS Machine Learning - Speciality certification. However, as of late August 2024, AWS has made the AWS Certified AI Practitioner certification available, and it’s currently in beta. As a foundational certification, this means the barrier to learning about AWS’s AI solutions is now far lower than before.
As someone who’s taken this certification and passed, this article is designed to help you do the same, and help you know what to expect should you sit the AIF-C01 exam.
Who is the AWS Certified AI Practitioner for?
The Certified AI Practitioner is a foundational-level certificate, and it’s perfect for anyone who’s dipping their toes into AWS’s various AI solutions. This could be IT managers or IT support professionals, or people in non-technical roles who want to understand these solutions, such as business analysts, product or project managers.
However, just because this exam is foundational does not mean you should underestimate it. The exam doesn’t cover topics deeply, but it does cover a lot of ground. By passing it, it demonstrates you have that broad knowledge of AI concepts and AWS services.
If you’ve had some exposure—say, up to six months—to these technologies, this exam will be a great one to take. It doesn’t go into the weeds with coding or data engineering, so you don’t need to be an expert in these to pass.
What the exam covers
The exam is structured around five key domains: Fundamentals of AI and ML, Fundamentals of Generative AI, Applications of Foundation Models, Guidelines for Responsible AI, and Security, Compliance, and Governance for AI Solutions.
1. Fundamentals of AI and ML
This section will make up approximately 20% of your exam. As the name suggests, it covers essential concepts like basic AI terms, concepts like neural networks, natural language processing, and the different learning methods such as supervised, unsupervised, and reinforcement learning.
There are also questions such as how AI can be practically used, such as assisting with decision making and automating tasks—all those things that otherwise eat up valuable time. On the flip side, it also tests you to identify where AI and ML solutions aren’t fit for purpose.
Expect questions around the capabilities of AWS’s range of AI and ML solutions, like:
2. Fundamentals of Generative AI
Next is Generative AI, a rapidly growing field, which makes up about a quarter of the exam’s scored content. For this part, you’ll need to be able to explain specific generative AI concepts. That includes things like prompt engineering, transformer-based models, and the lifecycle of a generative AI model, from pre-training all the way to deployment.
Just like with the last section, expect questions about the advantages and disadvantages of generative AI, and about what solutions and features AWS offers for developing GenAI applications. Expect questions on:
3. Applications of Foundation Models
Right now, this section makes up the largest portion of the exam at 28% of scored content. You definitely should learn about Foundational Models, or FMs, going in. FMs are pre-trained models that can be fine tuned for specific tasks, rather than other models that you might get out of the box and are only fit for general applications.
Knowing how to choose the right model and apply effective prompt engineering techniques will give you a significant advantage not just in passing this exam, but also in real-world scenarios beyond it.
Expect questions around the selection criteria you’d use to choose a pre-trained model such as cost, latency, and complexity, and services you might use.
4. Guidelines for Responsible AI
With great power comes great responsibility. Sound familiar? Like spider powers, AI is no different. This domain makes up a smaller fraction of the exam at 14%, but is still very important.
Expect questions around responsible AI such as bias detection, fairness, and the ethical implications of AI. Understanding these concepts helps make sure the AI solutions you’ll help make in the future are not only effective, but also fair and inclusive.
5. Security, Compliance, and Governance for AI Solutions.
The final domain is also worth about 14% of your score, and covers all the AWS solutions you can use to secure those future AI solutions. This covers:
- IAM roles, policies and permissions
- Encryption
- Amazon Macie
- AWS PrivateLink
- The AWS Shared Responsibility Model
For a full breakdown of all the things the exam can cover, be sure to check out the AWS exam guide.
New question formats
Historically, AWS exams have had two kinds of questions:
- Multiple choice, where there’s one correct response and three incorrect answers, called distractors, and
- Multiple response, where there’s two or more correct answers out of five response options, and you need to select all the correct responses to get credit for the question.
However, with this new exam, AWS has added three new question types:
- Ordering, where you get a list of 3 to 5 responses to complete a specific task, and you need to select the right ones and place them in the correct order.
- Matching, where you get a list of responses to match to 3 to 7 prompts. And
- Case study, where you are asked two or more questions about a scenario. In this situation, each question is scored individually.
Now, despite all this, we were only given multiple-choice and multiple-response when taking the exam at the time of this video. However, you should definitely be aware of these question types, as they’re in the AWS exam guide.
Conclusion
Now, as I said, this exam is quite broad, and covers more than I’ve mentioned here. On the surface, it might already sound like a lot. But with the right preparation, I’m confident that you’ll be able to pass this exam on your first try.
