Curriculum Overview: Identifying Common AI Workloads (AI-900)
Identify features of common AI workloads
Curriculum Overview: Identifying Common AI Workloads
This curriculum provides a foundational understanding of the primary artificial intelligence (AI) workloads as defined in the Microsoft Azure AI Fundamentals (AI-900) certification. It explores the distinct capabilities of Computer Vision, Natural Language Processing, Document Intelligence, and the emerging field of Generative AI.
Prerequisites
Before engaging with this curriculum, learners should possess the following:
- General IT Knowledge: Basic understanding of cloud computing concepts (SaaS, PaaS, IaaS).
- Data Literacy: Familiarity with the idea that data can be structured (tables) or unstructured (images, text).
- No Coding Required: While helpful, programming experience is not strictly necessary for this foundational overview.
Module Breakdown
The curriculum is structured into four core pillars representing the most common industry AI workloads.
| Module | Workload Category | Primary Focus | Difficulty |
|---|---|---|---|
| 1 | Computer Vision | Analyzing visual input (images, video, spatial data) | Intermediate |
| 2 | Natural Language Processing (NLP) | Understanding, interpreting, and generating human language | Intermediate |
| 3 | Document Intelligence | Extracting structured data from unstructured documents | Foundational |
| 4 | Generative AI | Creating new original content (text, image, code) | Advanced |
AI Workload Taxonomy
Learning Objectives per Module
Module 1: Computer Vision
- Identify Image Classification: Determining the primary subject of an image (e.g., "This is a cat").
- Identify Object Detection: Locating specific items within an image and providing coordinates (e.g., "There is a dog at [X,Y]").
- Identify Facial Analysis: Detecting human faces and analyzing attributes like age, emotion, or identity.
- Identify OCR: Recognizing and extracting printed or handwritten text from images.
Module 2: Natural Language Processing (NLP)
- Key Phrase Extraction: Identifying the main talking points in a document.
- Entity Recognition: Categorizing text into known groups (People, Places, Organizations).
- Sentiment Analysis: Quantifying the emotional tone (Positive, Negative, Neutral) of text.
- Translation & Speech: Converting text between languages and converting spoken word to text (and vice-versa).
Module 3: Document Intelligence
- Form Processing: Automating the extraction of data from receipts, invoices, and business forms.
- Knowledge Mining: Creating a searchable index of information from a massive library of unstructured data.
Module 4: Generative AI
- Content Creation: Generating human-like responses, marketing copy, or creative stories.
- Code Generation: Using natural language prompts to write programming code.
- Image Generation: Creating synthetic imagery based on text descriptions (DALL-E style).
Success Metrics
To demonstrate mastery of this curriculum, the learner should be able to:
- Match Scenario to Service: Given a business problem (e.g., "We need to read license plates"), correctly identify the workload (Computer Vision/OCR).
- Identify Responsible AI Alignment: Explain how each workload must adhere to principles of Fairness, Reliability, and Transparency.
- Differentiate ML vs. Generative AI: Distinguish between a model that predicts a value (Regression) and a model that creates a new paragraph (Generative AI).
- Exam Readiness: Achieve a consistent 80%+ score on practice assessments regarding AI Workload Identification (representing 15-20% of the AI-900 exam).
Real-World Application
AI workloads are not theoretical; they power the modern digital economy. Below is a conceptual model of how AI transforms raw data into business value.
Data Transformation Flow
\begin{tikzpicture}[node distance=2.5cm, every node/.style={fill=white, font=\small}, align=center] % Nodes \node (input) [draw, rectangle, rounded corners] {\textbf{Input Data}\ (Photos, Audio, PDF)}; \node (process) [draw, circle, right of=input, xshift=1.5cm, fill=blue!10] {\textbf{AI Workload}\ (CV, NLP, GenAI)}; \node (output) [draw, rectangle, rounded corners, right of=process, xshift=1.5cm] {\textbf{Business Insight}\ (Safety Alerts, Chatbots)};
% Arrows \draw [->, thick] (input) -- (process); \draw [->, thick] (process) -- (output);
% Labels \node at (2,-1) {\textit{Feature Extraction}}; \node at (6,-1) {\textit{Decision Support}}; \end{tikzpicture}
Industry Examples
- Retail: Using Computer Vision to enable "checkout-free" stores where cameras track items in a basket.
- Customer Service: Using NLP to analyze call center recordings for customer frustration levels (Sentiment Analysis).
- Healthcare: Using Document Intelligence to digitize decades of handwritten patient records into a searchable database.
- Software Development: Using Generative AI to suggest boilerplate code, accelerating the development lifecycle.
[!IMPORTANT] Throughout all these workloads, Responsible AI remains the cornerstone. Every implementation must be checked for bias (Fairness) and data protection (Privacy & Security).