I earned the Amazon Web Services (AWS) Certified AI Practitioner cert this winter through Nyla’s Continuous Learning benefit program, which provides employees with up to $5000 per year to use for classes, conferences, certifications, or dedicated study hours. This exam costs $100 and can be taken in an on-demand, remote fashion. This certification provides an opportunity to demonstrate proficiency in AI, ML and generative AI concepts and their applications. The exam covers both these concepts in general and their respective capabilities and tools within AWS’ suite of offerings.
The exam provides you with 90 minutes to complete 65 multiple choice questions. The exam has several focus areas:
• Domain 1: Fundamentals of AI and ML (20% of scored content)
• Domain 2: Fundamentals of Generative AI (24% of scored content)
• Domain 3: Applications of Foundation Models (28% of scored content)
• Domain 4: Guidelines for Responsible AI (14% of scored content)
• Domain 5: Security, Compliance, and Governance for AI Solutions (14% of
scored content)
Motivation
I pursued this certification due to the buzz and customer demand around Generative AI and the relevance and credibility of AWS. While my work experience with AWS was somewhat limited on recent past projects, it seems likely that my current project will leverage AWS and I’ll have the opportunity to architect solutions with AWS. Beyond my own current project, there’s the obvious ubiquitous nature of AWS as a cloud-industrybehemoth that makes this certification worthwhile long term. My goal was to gain a thorough understanding of all the AI, ML, and Gen AI capabilities within AWS in order to be able to pair potential customer use-cases with the proper AWS tool(s) to get the job done. In addition, I wanted to walk away with a clear approach for guiding government stakeholders through the dizzying web of terms in this topic area and all of the hype surrounding Gen AI; to distill it all down and frame for them which mission sets need Gen AI vs traditional ML or simply basic analysis.
Key Takeaways
AWS offers an impressive line of offerings for ML and Gen AI, particularly several low/no-code capabilities that require limited additional customization before they are ready-to-use for one’s use-cases. Some examples include Comprehend (for Natural Language Processing), Transcribe for audio to text, Rekognition for computer vision and Sagemaker Canvas for no-code ML solutions. Amazon Bedrock provides powerful foundational models as a basis for building generative AI applications. Sagemaker Jumpstart is another tool that rapidly accelerates deploying ML models. The course also helped me think through some of the trade offs within the field that need to be considered and articulated to stakeholders at the initial stages of designing solutions. More complex models tend to be more accurate, but they often lead to ‘black box’ solutions that are difficult to explain to client leadership and users, whereas more explainable approaches like a Decision-Tree are often less accurate. Another trade off is that while models with more parameters are more accurate, they are also more expensive and require more time to train. The same can be said for Gen AI solutions that that require many input and output tokens. Lastly is the trade-off between Gen AI’s efficiency and creativity and its risky potential to generate inaccurate or harmful content and expose sensitive information. The certification covers several measures that can create guardrails to protect against these common downsides of Gen AI and provides methods for refining models and solutions tailored for a stakeholder’s data and specific context.
Study Prep
I recommend the following study resources. This was the path that worked for me.
■ AWS Skillbuilder free companion course
● https://explore.skillbuilder.aws/learn/course/internal/view/elearning/19554/exam-prep-standard-course-aws-certified-ai-practitioner-aif-c01
■ Via Udemy, I completed Stephane Maarek’s prep course and practice exams
■ It is not necessary to understand the “buttonology” of using AWS solutions or have an AWS cloud account to prepare for this exam.
Conclusion
I recommend this certification both for the relevance of AI, ML, Gen AI and the AWS platform. While I went in concerned that the content would overly focus on AWS-specific material and be an expensive advertisement for Amazon, I found the majority of the content focused on the general tradecraft of AI/ML. Most of the emphasized best practices and concepts apply to the field universally. I felt I came away better equipped to guide conversations with customer leadership to talk them through the risks, advantages, hype, paranoia and proper use of Gen AI. In addition this course prepared me for designing and implementing solutions within AWS.
Certificate Link: https://www.credly.com/badges/c84f3e9f-e683-4b56-bca2-2907849c7ed0/public_url
