Earlier this year I passed both the AI-900: Microsoft Azure AI Fundamentals and the more advanced AI-102: Microsoft Azure AI Engineer Associate exams. This track was compelling dur to both the topical nature of the subject matter and as stepping stones toward my longer term plan of attaining AZ-305: Azure Solutions Architect Expert.

I found them to be quite different in scope, depth, and the type of preparation required. This short blog goes through some of these aspects and is intended as a primer to anyone else considering the certifications themselves.

AI-900: Easy Introduction

AI-900 was relatively straightforward. It offered a broad overview of AI concepts—like machine learning, natural language processing, and computer vision—without requiring any coding or architectural deep-dives. Coming from a technical background with some previous machine learning experience, I found most of it intuitive, and the questions leaned heavily on comprehension of high-level concepts rather than hands-on implementation.

This exam required only light study, mainly brushing up on Microsoft’s terminology and service offerings like Azure Cognitive Services.

Easy bits:

  • Conceptual questions about AI use cases.
  • General understanding of services like Language and Vision.
  • No need to know how to deploy or build anything.

Challenging bits:

  • Minimal, though I had to adjust to how Microsoft frames certain scenarios in their own language.

AI-102: A Step Up in Complexity

AI-102, on the other hand, required much deeper understanding, especially around building, securing, and optimising AI solutions on Azure. This included working with the SDKs, REST APIs, and designing real-life solutions involving many of the services that come under the Azure AI Services umbrella.

My background as a PhD and data scientist helped with grasping architecture patterns and integration scenarios, but the coding-level detail—especially around authoring LUIS models, using the Form Recognizer, or configuring QnA Maker—demanded a more focused, technical preparation that instead drew from my software engineering experience.

Easy bits (due to experience):

  • Designing end-to-end solutions and securing APIs.
  • Understanding architectural patterns for deploying AI workloads.
  • Aligning AI services with enterprise use-cases.

Challenging bits:

  • Low-level configuration, SDK and REST API details.
  • Precise syntax or steps for integrating services like Speech or Language Understanding.
  • Edge cases in deploying and optimizing AI models.

Summary of Study Effort

  • AI-900: ~1 week of light review, ideal for non-technical or semi-technical professionals.
  • AI-102: 3–4 weeks of deeper study, especially if you’re not hands-on with Azure AI services daily. Hands-on labs and practice tests were essential.

Bonus: Simulate Exam Questions with AI

The irony of practices for AI exams using AI itself is not lost on me, however, I would say it works very well and saves on the need to track down (and purchase in many cases) exam questions. Instead I use the following prompt to have ChatGPT continually test me for my chosen exam:

I am an IT Consultant working as a Solution Architect in financial services institutions. I am looking for factual and punchy study assistance towards relevant Azure certifications. My background is in resaarch, development, cloud and R&D.

I require question simulations for AI-102; I would like hard questions that are representative of the requested certification, one question at a time, with the result presented right away as correct or incorrect with a tick or a cross; along with an explanation of the answers. Keep the questions coming automatically (don’t ask) with checkpoints every 10 questions to show progress and if I am tracking as exam ready. Make sure the questions are a cross-section of all the subjects in the given exam.

🧑‍🎓 Happy studying! 🧑‍🎓