azure-ai-900-prep-notes

📘 Study notes, key concepts, and learning resources for Microsoft Azure AI-900: Azure AI Fundamentals certification. Includes personal takeaways, visual breakdowns, and LinkedIn-ready materials.

🧠 AZURE AI FUNDAMENTALS – DETAILED STUDY NOTES 🔷 1. Azure AI Services – Overview Azure AI Services is a collective term for Microsoft’s AI tools and platforms. It includes:

Azure Cognitive Services (pre-built AI)

Azure Machine Learning (build/train custom models)

Azure OpenAI Service (LLMs like GPT, DALL·E, Whisper)

Azure AI Studio (unified GenAI platform)

Azure Bot Service (Conversational AI)

🔷 2. Azure Cognitive Services – Categories & Details A. 🔍 Vision Pre-trained APIs to analyze visual content.

Service Description Use Cases Computer Vision Analyze images (objects, tags, description, OCR) Detect labels, get alt-text Custom Vision Train your own image classifier Classify damaged vs normal products Face API Detect faces, attributes (age, emotion) Retail insights, security checks Form Recognizer Extract fields, tables from docs Invoices, tax forms, IDs Semantic Segmentation Pixel-level classification of objects Satellite imagery, medical scans Domain Models Celebrities, Landmarks Media tagging, tourism apps

B. 🗣️ Speech Convert between speech and text.

Service Description Use Cases Speech to Text Audio → Text Meeting transcripts Text to Speech Text → Voice Read out articles, IVR Speech Translation Live speech translation Multilingual conferences Speaker Recognition Identify/verify speaker Voice login, call centers

C. 🌐 Language Natural Language Processing (NLP) services.

Service Description Use Cases Text Analytics Extract key phrases, sentiment, PII, entities Analyze feedback, redact data Translator (NMT) Translate text in 90+ languages Chatbots, web content Language Understanding (LUIS) Detect user intents/entities Chatbots: "Book a flight" Question Answering (ex-QnA Maker) FAQ-style Q&A from documents Self-service portals

D. 🤖 Decision Services for intelligent recommendations and moderation.

Service Description Use Cases Personalizer Reinforcement learning-based personalization News, product suggestions Content Moderator Detect offensive content in text/images Social media platforms Anomaly Detector Detect time-series anomalies Server monitoring, fraud detection

🔷 3. Azure Machine Learning – Detailed View Used to build, train, and deploy custom ML models.

Component Description Keywords Designer Drag-drop UI for building ML pipelines No code AutoML Auto-select algorithm + hyperparameters Classification, regression Notebooks Jupyter-style coding for data scientists Python Datasets Managed training data Reusable Compute Targets CPU/GPU VMs for training Scale up Model Registry Version control for ML models Track performance Endpoints Deploy model as REST API Real-time prediction Pipelines Automate end-to-end ML workflows Production-ready

📌 ML Lifecycle:

Data → Preprocessing → Feature Engineering

Split into Train/Test

Train model → Evaluate (Accuracy, F1, AUC)

Deploy → Monitor

🔷 4. Machine Learning Concepts Term Definition Examples Supervised Learning Input + known output Classification, regression Unsupervised Learning No output labels Clustering Classification Predict label/class Email: spam/not spam Regression Predict numeric value Price prediction Clustering Group similar items Customer segmentation PCA Reduce features Visualization Ordinal Regression Predict ranked categories Ratings (1–5) Poisson Regression Predict counts # of calls per day

🔷 5. Evaluation Metrics Metric Purpose When to Use Accuracy Overall correctness Balanced data Precision Correct Positives / All Positives Reduce false positives Recall Correct Positives / All Actual Positives Reduce false negatives F1 Score Harmonic mean of precision & recall Imbalanced classes AUC Class separability Binary classifiers

🔷 6. Responsible AI Principles Principle Description Fairness Prevent bias, ensure equity Inclusiveness Accessible to all users Reliability & Safety Work as intended in all scenarios Transparency Explain how decisions are made Accountability Human oversight and responsibility Privacy & Security Protect personal and sensitive data

🔷 7. Knowledge Mining Combines OCR, Form Recognizer, Text Analytics, and Azure Cognitive Search

Used to turn unstructured documents into searchable knowledge

Common in legal, healthcare, finance

🔷 8. Conversational AI & Bot Service Azure Bot Service + Cognitive Services

Tools:

Bot Framework SDK

LUIS for intent recognition

QnA for FAQ-style bots

Channels: Web Chat, MS Teams, Telegram, WhatsApp

🔷 9. Azure OpenAI & AI Studio (Modern AI) Tool Purpose Azure AI Studio Unified GUI for GenAI + prompt flow Prompt Flow LLM pipeline (prompt → completion) Azure OpenAI Service GPT-4, Codex, DALL·E, Whisper Data Grounding RAG – connect model to private data Fine-Tuning Customize LLMs with your data

🔷 10. Modern Naming Map (AI Studio) Exam Term Old Name New Name Vision Service Computer Vision Azure AI Vision – Image Analysis OCR Read API Azure AI Vision OCR Text Analytics Text Analytics Azure AI Content Understanding Form Recognizer Form Recognizer Azure AI Document Intelligence LUIS Language Understanding Conversational Language Understanding QnA Maker QnA Maker Azure Language – Question Answering Translator Translator Azure Language – Text Translation

🔷 11. Final Tips Focus on use cases + keywords

Know which service fits which scenario

Practice drag-drop scenario questions

Understand end-to-end ML workflows