Functional Group
A section of the exam covering related skills.
Getting Started: Your AI-102 Journey
Free knowledge base
Microsoft Certified: Azure AI Engineer Associate — key terms, tricks & tips
Everything from the course in one searchable place: 277 entries. Use it to review before a practice test or look up a word you forgot.
277 results
Weighting
Percentage indicating importance of a domain.
Getting Started: Your AI-102 Journey
Lab Question
Practical task in a simulated Azure environment.
Getting Started: Your AI-102 Journey
Skills Measured
Official document outlining exam topics.
Getting Started: Your AI-102 Journey
Microsoft Learn
Microsoft's free online learning platform.
Getting Started: Your AI-102 Journey
Associate-level
Certification for professionals with practical experience.
Getting Started: Your AI-102 Journey
Understanding the AI-102 Exam Structure
To remember the key steps for exam prep: 'R.S.H.P.I.R.' - Review, Study, Hands-on, Practice, Identify, Re-study.
Getting Started: Your AI-102 Journey
Understanding the AI-102 Exam Structure
Memorize the approximate weighting percentages for each functional group. While exact numbers can shift, understanding which domains are most heavily tested is critical for exam strategy. Always check the official 'Skills Measured' document before your exam date.
Getting Started: Your AI-102 Journey
Understanding the AI-102 Exam Structure
Only studying theoretical concepts and neglecting hands-on practice, especially for lab questions.
Getting Started: Your AI-102 Journey
Understanding the AI-102 Exam Structure
Not checking the official 'Skills Measured' document for the most current exam objectives and weightings.
Getting Started: Your AI-102 Journey
Understanding the AI-102 Exam Structure
Ignoring time management during practice exams, leading to rushed answers on the actual test.
Getting Started: Your AI-102 Journey
Azure Machine Learning
Cloud-based service for end-to-end ML lifecycle management.
Getting Started: Your AI-102 Journey
Workspace
Top-level Azure ML resource, a central hub for ML project artifacts.
Getting Started: Your AI-102 Journey
Resource Group
Logical container for Azure resources.
Getting Started: Your AI-102 Journey
Azure Cloud Shell
Browser-based command-line interface for Azure management.
Getting Started: Your AI-102 Journey
Azure Storage Account
Service for storing various types of data.
Getting Started: Your AI-102 Journey
Azure Key Vault
Service for securely storing secrets, keys, and certificates.
Getting Started: Your AI-102 Journey
Azure Container Registry
Service for storing and managing Docker container images.
Getting Started: Your AI-102 Journey
Compute Instance
Cloud-based workstation for development within Azure ML.
Getting Started: Your AI-102 Journey
Setting Up Your Azure AI Development Environment
To remember the four automatic dependencies of an Azure ML Workspace, think: 'S-K-A-C' for Storage, Key Vault, Application Insights, Container Registry. Sounds like 'Stack'!
Getting Started: Your AI-102 Journey
Setting Up Your Azure AI Development Environment
The exam often tests your knowledge of the core components provisioned with an Azure ML workspace: Storage Account, Key Vault, Application Insights, and Container Registry. Memorize these four dependencies.
Getting Started: Your AI-102 Journey
Setting Up Your Azure AI Development Environment
Forgetting to specify a unique name for your Azure ML workspace, leading to deployment failures.
Getting Started: Your AI-102 Journey
Setting Up Your Azure AI Development Environment
Not understanding that an Azure ML workspace automatically provisions other core Azure services, leading to confusion during troubleshooting.
Getting Started: Your AI-102 Journey
Setting Up Your Azure AI Development Environment
Attempting to deploy models without having an Azure Container Registry linked or available for image storage.
Getting Started: Your AI-102 Journey
Azure Cognitive Services
Cloud-based APIs for pre-built AI capabilities like vision, speech, language.
Planning & Managing Azure AI Solutions
Azure Applied AI Services
Higher-level services combining AI capabilities for specific business problems.
Planning & Managing Azure AI Solutions
Text Analytics
Cognitive Service for extracting sentiment, key phrases, and entities from text.
Planning & Managing Azure AI Solutions
Form Recognizer
Applied AI Service for extracting information from documents and forms.
Planning & Managing Azure AI Solutions
Computer Vision
Cognitive Service for image analysis, object detection, and optical character recognition.
Planning & Managing Azure AI Solutions
MLOps
Practices for reliable and efficient deployment and maintenance of ML models.
Planning & Managing Azure AI Solutions
Selecting the Right Azure AI Service
To remember the service categories, think: 'C.A.M. you pick the right AI?' C is for Cognitive, A is for Applied, M is for Machine Learning.
Planning & Managing Azure AI Solutions
Selecting the Right Azure AI Service
The exam frequently presents scenarios and asks you to choose the BEST Azure AI service. Look for keywords like 'pre-built capabilities,' 'common tasks,' or 'quick integration' for Cognitive Services. For 'end-to-end solution' or 'specific business problem,' consider Applied AI Services. If the scenario mentions 'custom models,' 'unique data,' or 'fine-grained control,' Azure Machine Learning is the answer.
Planning & Managing Azure AI Solutions
Selecting the Right Azure AI Service
Over-engineering by choosing Azure Machine Learning for a problem that could be solved with a simpler Cognitive Service, leading to increased development time and cost.
Planning & Managing Azure AI Solutions
Selecting the Right Azure AI Service
Trying to force a Cognitive Service to perform a highly specialized task it wasn't designed for, resulting in poor accuracy and complex workarounds.
Planning & Managing Azure AI Solutions
Selecting the Right Azure AI Service
Not considering the cost implications of different services; pre-built services often have a pay-per-use model, while custom ML can incur significant compute costs.
Planning & Managing Azure AI Solutions
Shared Responsibility Model
Defines security duties between cloud provider and customer.
Planning & Managing Azure AI Solutions
Azure RBAC
Role-Based Access Control; manages permissions to Azure resources.
Planning & Managing Azure AI Solutions
Customer-Managed Keys (CMK)
Encryption keys managed by the customer for data at rest.
Planning & Managing Azure AI Solutions
Adversarial Attack
Malicious input designed to trick an AI model.
Planning & Managing Azure AI Solutions
Data Poisoning
Manipulating training data to compromise model integrity.
Planning & Managing Azure AI Solutions
Azure Monitor
Service for collecting and analyzing telemetry data.
Planning & Managing Azure AI Solutions
Log Analytics
Azure Monitor component for querying and analyzing log data.
Planning & Managing Azure AI Solutions
Azure Policy
Enforces organizational standards and assesses compliance.
Planning & Managing Azure AI Solutions
Securing and Monitoring Azure AI Services
For AI SECURE: S-hared responsibility, E-ncryption, C-ompliance, U-ser access, R-obust monitoring, E-volve defenses.
Planning & Managing Azure AI Solutions
Securing and Monitoring Azure AI Services
The exam frequently tests on the shared responsibility model. Remember: Microsoft secures the cloud, you secure in the cloud. Also, know that Azure Monitor is the primary tool for monitoring, including metrics, logs, and alerts.
Planning & Managing Azure AI Solutions
Securing and Monitoring Azure AI Services
Forgetting to implement RBAC, leaving AI service endpoints publicly accessible without proper authentication.
Planning & Managing Azure AI Solutions
Securing and Monitoring Azure AI Services
Neglecting data encryption for sensitive data, especially when using custom models or bringing your own data.
Planning & Managing Azure AI Solutions
Securing and Monitoring Azure AI Services
Not setting up alerts for critical metrics, leading to delayed detection of performance issues or security breaches.
Planning & Managing Azure AI Solutions
Consumption-based billing
Paying for cloud resources based on actual usage.
Planning & Managing Azure AI Solutions
Reserved Instances
Pre-purchasing compute capacity for cost savings.
Planning & Managing Azure AI Solutions
Azure Cost Management
Azure service for tracking and optimizing cloud spending.
Planning & Managing Azure AI Solutions
Cloud Deployment
Running AI services entirely within Azure's data centers.
Planning & Managing Azure AI Solutions
Edge Deployment
Running AI models on devices closer to data sources.
Planning & Managing Azure AI Solutions
Hybrid Deployment
Combining cloud and on-premises/edge AI components.
Planning & Managing Azure AI Solutions
Azure IoT Edge
Managed service for deploying cloud workloads to edge devices.
Planning & Managing Azure AI Solutions
Cost Management and Deployment Strategies for AI
To manage AI costs and deployments, remember 'C.O.D.E.': **C**osts (monitor!), **O**ptimize (resources!), **D**eploy (cloud/edge/hybrid!), **E**valuate (requirements!).
Planning & Managing Azure AI Solutions
Cost Management and Deployment Strategies for AI
Memorize that Azure AI services are primarily billed on a consumption model. For the exam, recognize keywords like 'latency-sensitive' or 'disconnected environment' as indicators for edge deployment, and 'scalability' or 'managed service' for cloud deployment.
Planning & Managing Azure AI Solutions
Cost Management and Deployment Strategies for AI
Over-provisioning resources without considering actual usage patterns, leading to unnecessary costs.
Planning & Managing Azure AI Solutions
Cost Management and Deployment Strategies for AI
Ignoring Azure Cost Management tools and not regularly reviewing spending, missing opportunities for optimization.
Planning & Managing Azure AI Solutions
Cost Management and Deployment Strategies for AI
Choosing a cloud-only deployment for applications with strict low-latency requirements, resulting in poor user experience.
Planning & Managing Azure AI Solutions
Scalability
Ability of a system to handle increasing workload without performance degradation.
Planning & Managing Azure AI Solutions
Resilience
Ability of a system to recover from failures and continue functioning.
Planning & Managing Azure AI Solutions
High Availability (HA)
Ensuring continuous operation with minimal downtime, often via redundancy.
Planning & Managing Azure AI Solutions
Disaster Recovery (DR)
Strategies to restore services after a major outage, often cross-region.
Planning & Managing Azure AI Solutions
Availability Zones
Physically separate data centers within an Azure region for HA.
Planning & Managing Azure AI Solutions
Retry Logic
Mechanism to re-attempt failed operations, often with exponential backoff.
Planning & Managing Azure AI Solutions
Circuit Breaker
Pattern to prevent repeated calls to a failing service, preventing cascading failures.
Planning & Managing Azure AI Solutions
Fan-out/Fan-in
Pattern for parallel processing where tasks are distributed and results aggregated.
Planning & Managing Azure AI Solutions
Designing Scalable and Resilient AI Solutions
Remember 'S.R. H.A. D.R.' for Scalability, Resilience, High Availability, Disaster Recovery – the core pillars of robust AI.
Planning & Managing Azure AI Solutions
Designing Scalable and Resilient AI Solutions
The exam frequently tests your understanding of multi-region deployments for disaster recovery and the use of Availability Zones for high availability within a region. Know the difference and when to apply each.
Planning & Managing Azure AI Solutions
Designing Scalable and Resilient AI Solutions
Confusing High Availability (HA) with Disaster Recovery (DR); HA is within a region, DR is across regions.
Planning & Managing Azure AI Solutions
Designing Scalable and Resilient AI Solutions
Not implementing retry logic or circuit breakers for transient faults, leading to brittle applications.
Planning & Managing Azure AI Solutions
Designing Scalable and Resilient AI Solutions
Failing to regularly test disaster recovery plans, assuming they will work when needed.
Planning & Managing Azure AI Solutions
Application Insights
APM for live web apps, detects performance anomalies.
Planning & Managing Azure AI Solutions
Kusto Query Language (KQL)
Powerful language for querying data in Log Analytics.
Planning & Managing Azure AI Solutions
Action Group
Defines what happens when an Azure alert fires.
Planning & Managing Azure AI Solutions
Metrics
Numerical values describing system performance over time.
Planning & Managing Azure AI Solutions
Logs
Timestamped events recording activity in a system.
Planning & Managing Azure AI Solutions
Implementing Azure AI Service Monitoring
MALTA: Metrics, Alerts, Logs, Traces, Actions. Remember these five pillars for comprehensive Azure AI monitoring!
Planning & Managing Azure AI Solutions
Implementing Azure AI Service Monitoring
For the AI-102 exam, understand that Azure Monitor is the primary tool for all monitoring, and Application Insights is its specialized component for application performance. Be ready to distinguish between metrics (numerical data) and logs (event records) and how KQL is used with logs.
Planning & Managing Azure AI Solutions
Implementing Azure AI Service Monitoring
Ignoring non-critical alerts: Even low-priority alerts can indicate a developing problem that could escalate.
Planning & Managing Azure AI Solutions
Implementing Azure AI Service Monitoring
Not configuring action groups: Alerts without corresponding actions are useless; ensure notifications reach the right team.
Planning & Managing Azure AI Solutions
Implementing Azure AI Service Monitoring
Over-monitoring or under-monitoring: Find the right balance; too many alerts lead to alert fatigue, too few mean missed issues.
Planning & Managing Azure AI Solutions
Azure AI Vision
Cloud service for image analysis, processing, and understanding.
Image and Video Processing with Azure AI
Image Classification
Assigning labels or categories to an entire image's content.
Image and Video Processing with Azure AI
Object Detection
Identifying and locating specific objects within an image.
Image and Video Processing with Azure AI
Facial Recognition
Detecting human faces and analyzing their attributes or identity.
Image and Video Processing with Azure AI
Content Moderation
Automatically identifying and filtering inappropriate image content.
Image and Video Processing with Azure AI
Bounding Box
Rectangular coordinates indicating an object's location in an image.
Image and Video Processing with Azure AI
Smart Thumbnail
Intelligently cropped image focusing on the most important region.
Image and Video Processing with Azure AI
Captioning
Generating a descriptive text summary for an image's content.
Image and Video Processing with Azure AI
Analyzing Images with Azure AI Vision
Imagine a 'VISION' test: V for Visual analysis, I for Insights, S for Smart features, I for Integration, O for Object detection, N for Naughty content moderation.
Image and Video Processing with Azure AI
Analyzing Images with Azure AI Vision
The exam often tests your understanding of specific Azure AI Vision capabilities. Be ready to differentiate between image classification (what is it?) and object detection (where is it?). Also, know that content moderation is a key use case for filtering inappropriate content.
Image and Video Processing with Azure AI
Analyzing Images with Azure AI Vision
Confusing image classification (what the image is about) with object detection (where specific items are located).
Image and Video Processing with Azure AI
Analyzing Images with Azure AI Vision
Assuming Azure AI Vision automatically performs all possible analyses; you must specify the desired features in your API call.
Image and Video Processing with Azure AI
Analyzing Images with Azure AI Vision
Forgetting that content moderation is a distinct and configurable feature, not just a byproduct of general analysis.
Image and Video Processing with Azure AI
OCR
Optical Character Recognition; extracts text from images.
Image and Video Processing with Azure AI
Read API
Advanced Azure AI Vision OCR for documents and images.
Image and Video Processing with Azure AI
Synchronous OCR
API returns results immediately for small images.
Image and Video Processing with Azure AI
Asynchronous OCR
API returns operation ID; poll for results later for large documents.
Image and Video Processing with Azure AI
Confidence Score
A value indicating the certainty of the OCR extraction.
Image and Video Processing with Azure AI
Handwritten Text Recognition
OCR capability to extract text written by hand.
Image and Video Processing with Azure AI
Extracting Text with Optical Character Recognition (OCR)
Think of OCR as 'Oh, Can Read!' – it's like teaching your computer to read any text it sees, whether it's printed on a sign or scribbled in a note.
Image and Video Processing with Azure AI
Extracting Text with Optical Character Recognition (OCR)
For the exam, remember that the Read API is Microsoft's recommended OCR solution for general-purpose text extraction from documents and images, especially for multi-page documents and handwritten text. Keywords like 'document processing,' 'multi-page PDF,' or 'handwritten' should immediately trigger 'Read API' in your mind. Also, distinguish between synchronous (quick, small) and asynchronous (long-running, large) operations.
Image and Video Processing with Azure AI
Extracting Text with Optical Character Recognition (OCR)
Using synchronous OCR for large, multi-page documents, leading to timeouts.
Image and Video Processing with Azure AI
Extracting Text with Optical Character Recognition (OCR)
Not pre-processing images (e.g., de-skewing, enhancing contrast) before sending them to OCR, resulting in lower accuracy.
Image and Video Processing with Azure AI
Extracting Text with Optical Character Recognition (OCR)
Ignoring confidence scores and blindly accepting all extracted text without human review for critical data.
Image and Video Processing with Azure AI
Custom Vision
Azure service for building custom image classification and object detection models.
Image and Video Processing with Azure AI
Training Data
Labeled images used to teach a machine learning model to recognize patterns.
Image and Video Processing with Azure AI
Precision
Proportion of positive identifications that were actually correct.
Image and Video Processing with Azure AI
Recall
Proportion of actual positives that were correctly identified.
Image and Video Processing with Azure AI
mAP (Mean Average Precision)
Evaluation metric for object detection, averaging precision across classes.
Image and Video Processing with Azure AI
Building Custom Vision Models for Image Analysis
To remember the Custom Vision workflow: 'Upload, Tag, Train, Test, Deploy' – UTTTD, like 'Up To The Top, Down!'
Image and Video Processing with Azure AI
Building Custom Vision Models for Image Analysis
The exam often tests the difference between image classification and object detection, and the typical workflow steps for Custom Vision. Pay attention to the minimum recommended number of images per tag (generally 50) and the purpose of key metrics like precision, recall, and mAP.
Image and Video Processing with Azure AI
Building Custom Vision Models for Image Analysis
Not providing enough diverse training images, leading to a model that performs poorly on new, unseen data.
Image and Video Processing with Azure AI
Building Custom Vision Models for Image Analysis
Confusing image classification with object detection; classification tags the whole image, detection finds and localizes objects.
Image and Video Processing with Azure AI
Building Custom Vision Models for Image Analysis
Neglecting to iterate and improve the model by adding more data or correcting mislabeled images after initial training.
Image and Video Processing with Azure AI
Azure AI Video Indexer
Cloud service for extracting deep insights from video and audio content.
Image and Video Processing with Azure AI
Speaker Diarization
Identifying who spoke when in an audio recording or video.
Image and Video Processing with Azure AI
Sentiment Analysis
Determining the emotional tone (positive, negative, neutral) of spoken text.
Image and Video Processing with Azure AI
Visual Insights
Information extracted from the video stream, e.g., objects, faces, brands.
Image and Video Processing with Azure AI
Audio Insights
Information extracted from the audio stream, e.g., speech, sentiment, keywords.
Image and Video Processing with Azure AI
REST APIs
Programmatic interface for interacting with Video Indexer services.
Image and Video Processing with Azure AI
Processing and Analyzing Video Content
VIDEO: VIsual and audio, Detects objects, Emotions, Diarization, Objects.
Image and Video Processing with Azure AI
Processing and Analyzing Video Content
The exam often tests your understanding of Video Indexer's capabilities. Look for scenarios involving extracting multiple types of insights (speech, faces, brands, emotions) from video or audio, or automating content processing.
Image and Video Processing with Azure AI
Processing and Analyzing Video Content
Confusing Video Indexer with simple video encoding services; Video Indexer provides deep content understanding, not just format conversion.
Image and Video Processing with Azure AI
Processing and Analyzing Video Content
Underestimating the breadth of insights; remember it covers both audio (speech, sentiment) and visual (objects, faces, brands).
Image and Video Processing with Azure AI
Processing and Analyzing Video Content
Forgetting that it can be integrated via APIs, not just used through its portal.
Image and Video Processing with Azure AI
Named Entity Recognition (NER)
Identifies and categorizes entities like people, places, organizations.
Natural Language Processing on Azure
Key Phrase Extraction
Identifies the main concepts or topics within a document.
Natural Language Processing on Azure
PII Detection
Identifies and redacts sensitive Personally Identifiable Information.
Natural Language Processing on Azure
Text Summarization
Generates a concise summary of a longer text document.
Natural Language Processing on Azure
Language Detection
Automatically identifies the primary language of the input text.
Natural Language Processing on Azure
API Key
Credential used for authenticating requests to Azure AI services.
Natural Language Processing on Azure
Analyzing Text with Azure AI Language
S-K-N-P-L-T: **S**entiment, **K**ey phrases, **N**amed entities, **P**II, **L**anguage, **T**ext summarization. Remember these core features!
Natural Language Processing on Azure
Analyzing Text with Azure AI Language
For the AI-102 exam, understand the specific capabilities of each Azure AI Language feature. Keywords like 'emotional tone' point to Sentiment Analysis, 'main topics' to Key Phrase Extraction, and 'people, places, organizations' to Named Entity Recognition. Be ready to choose the correct feature for a given scenario.
Natural Language Processing on Azure
Analyzing Text with Azure AI Language
Confusing Key Phrase Extraction with Text Summarization; Key Phrase extracts important terms, Summarization creates a condensed version of the text.
Natural Language Processing on Azure
Analyzing Text with Azure AI Language
Not understanding the difference between Named Entity Recognition (NER) and custom entity recognition (which might require a custom model).
Natural Language Processing on Azure
Analyzing Text with Azure AI Language
Forgetting to secure API keys or hardcoding them directly into application code, which is a security risk.
Natural Language Processing on Azure
Speech-to-Text (STT)
Converts spoken audio into written text.
Natural Language Processing on Azure
Text-to-Speech (TTS)
Converts written text into spoken audio.
Natural Language Processing on Azure
Neural Voices
Highly natural, human-like synthetic voices.
Natural Language Processing on Azure
SSML
Markup language for controlling TTS output.
Natural Language Processing on Azure
Custom Speech
Training STT models with domain-specific data.
Natural Language Processing on Azure
Custom Neural Voice
Creating a unique, brand-specific TTS voice.
Natural Language Processing on Azure
Implementing Speech-to-Text and Text-to-Speech
STT: 'Speak To Text' – S for Speak, T for Text. TTS: 'Text To Sound' – T for Text, S for Sound.
Natural Language Processing on Azure
Implementing Speech-to-Text and Text-to-Speech
The exam often tests your understanding of when to use custom speech models versus standard models, and the difference between pre-built neural voices and custom neural voices. Look for keywords like 'domain-specific vocabulary' for custom speech, or 'brand identity' for custom neural voice.
Natural Language Processing on Azure
Implementing Speech-to-Text and Text-to-Speech
Confusing the purpose of custom speech models (improving STT accuracy for specific vocabulary) with custom neural voice (creating a unique TTS voice).
Natural Language Processing on Azure
Implementing Speech-to-Text and Text-to-Speech
Underestimating the importance of SSML for fine-tuning TTS output, assuming basic text input is always sufficient.
Natural Language Processing on Azure
Implementing Speech-to-Text and Text-to-Speech
Forgetting that a Speech resource in Azure is required for both STT and TTS functionalities, not separate resources.
Natural Language Processing on Azure
Entity
A specific, relevant piece of information extracted from an utterance.
Natural Language Processing on Azure
Utterance
The input text or spoken phrase provided by the user.
Natural Language Processing on Azure
LUIS
Language Understanding Intelligent Service, a legacy Azure NLU service.
Natural Language Processing on Azure
CLU
Conversational Language Understanding, the modern NLU service in Azure AI Language.
Natural Language Processing on Azure
Schema
The defined set of intents and entities in a language understanding project.
Natural Language Processing on Azure
Active Learning
Process of using real user utterances to improve model accuracy.
Natural Language Processing on Azure
Building Language Understanding Solutions (LUIS/CLU)
CLU is your 'Clever Language Understander' — it's smart, modern, and in the 'Language' family!
Natural Language Processing on Azure
Building Language Understanding Solutions (LUIS/CLU)
The exam emphasizes CLU as the preferred and modern service for new projects. Be prepared to explain why CLU is better than LUIS (e.g., better accuracy, multilingual, part of Azure AI Language). Keywords to spot: 'Conversational Language Understanding', 'Azure AI Language', 'intent recognition', 'entity extraction'.
Natural Language Processing on Azure
Building Language Understanding Solutions (LUIS/CLU)
Not providing enough diverse example utterances, leading to poor model accuracy.
Natural Language Processing on Azure
Building Language Understanding Solutions (LUIS/CLU)
Overlapping intent definitions, making it hard for the model to distinguish between similar user goals.
Natural Language Processing on Azure
Building Language Understanding Solutions (LUIS/CLU)
Forgetting to label all relevant entities in example utterances, reducing information extraction capabilities.
Natural Language Processing on Azure
Generative AI
AI models that create new, original content (text, images, etc.).
Natural Language Processing on Azure
Discriminative Model
AI models that classify or predict a label for given input data.
Natural Language Processing on Azure
Large Language Model (LLM)
Deep learning models with billions of parameters, trained on vast text data.
Natural Language Processing on Azure
Natural Language Processing (NLP)
A field of AI focused on enabling computers to understand, interpret, and generate human language.
Natural Language Processing on Azure
Transformer Architecture
A neural network architecture widely used in LLMs, known for its attention mechanism.
Natural Language Processing on Azure
Prompt Engineering
The art of crafting effective inputs (prompts) to guide Generative AI models.
Natural Language Processing on Azure
Introduction to Generative AI for NLP
GENERATE new ideas, DISCRIMINATE between old ones. LLMs are LARGE and can LEARN a LOT.
Natural Language Processing on Azure
Introduction to Generative AI for NLP
The exam expects you to differentiate between generative and discriminative AI models and understand the core capabilities and common applications of LLMs within Azure AI services, particularly Azure OpenAI Service.
Natural Language Processing on Azure
Introduction to Generative AI for NLP
Confusing generative models (create) with discriminative models (classify).
Natural Language Processing on Azure
Introduction to Generative AI for NLP
Underestimating the role of LLMs as the backbone of modern generative NLP.
Natural Language Processing on Azure
Introduction to Generative AI for NLP
Believing Generative AI is only for simple text generation, ignoring its broader applications like summarization or code generation.
Natural Language Processing on Azure
Retrieval Augmented Generation (RAG)
Combines LLM generation with external data retrieval for grounded responses.
Natural Language Processing on Azure
Fine-tuning
Further training a pre-trained LLM on a specific dataset to adapt its behavior.
Natural Language Processing on Azure
Hallucination
When an LLM generates plausible but factually incorrect or nonsensical information.
Natural Language Processing on Azure
Few-shot Learning
Prompt engineering technique providing a few examples to guide LLM behavior.
Natural Language Processing on Azure
Chain-of-Thought Prompting
Prompting technique asking LLM to show its reasoning steps before the final answer.
Natural Language Processing on Azure
Advanced Generative AI for Language Solutions
RAG: 'Retrieve And Generate' – you first get the facts, then you let the AI talk. Fine-tuning: 'Focused Training' – you teach the AI new tricks for specific tasks.
Natural Language Processing on Azure
Advanced Generative AI for Language Solutions
The exam often asks about the benefits and appropriate use cases for RAG vs. fine-tuning. Remember RAG is for grounding with external, dynamic data, while fine-tuning adapts the model's core knowledge and style.
Natural Language Processing on Azure
Advanced Generative AI for Language Solutions
Assuming LLMs are always factually correct without grounding or validation.
Natural Language Processing on Azure
Advanced Generative AI for Language Solutions
Using overly vague or ambiguous prompts, leading to irrelevant or unhelpful responses.
Natural Language Processing on Azure
Advanced Generative AI for Language Solutions
Applying fine-tuning when simple prompt engineering or RAG would suffice, incurring unnecessary cost and complexity.
Natural Language Processing on Azure
Azure AI Search
Managed cloud search service for rich search experiences.
Knowledge Mining with Azure AI
Indexer
Automates data ingestion from a data source into an index.
Knowledge Mining with Azure AI
Skillset
Collection of AI skills to enrich data before indexing.
Knowledge Mining with Azure AI
Search Index
The searchable data store within Azure AI Search.
Knowledge Mining with Azure AI
Semantic Search
Uses AI to understand query intent and context for relevance.
Knowledge Mining with Azure AI
Vector Search
Searches for semantically similar items using vector embeddings.
Knowledge Mining with Azure AI
Data Source
Connection to external data like Blob Storage or SQL DB.
Knowledge Mining with Azure AI
Query Language
Syntax used to retrieve information from a search index.
Knowledge Mining with Azure AI
Implementing Azure AI Search for Information Retrieval
To remember the core components: 'DISSI' - Data source, Indexer, Skillset, Search Index. Imagine a 'dizzy' process of getting data ready for search!
Knowledge Mining with Azure AI
Implementing Azure AI Search for Information Retrieval
The exam often tests your understanding of the core components (data source, indexer, skillset, index) and their purpose. Pay attention to scenarios where you'd use a skillset versus just an indexer, especially for unstructured data. Keywords: 'enrichment', 'unstructured content', 'natural language query'.
Knowledge Mining with Azure AI
Implementing Azure AI Search for Information Retrieval
Forgetting to run the indexer after making changes to the data source or skillset, leading to outdated search results.
Knowledge Mining with Azure AI
Implementing Azure AI Search for Information Retrieval
Not configuring appropriate access policies or API keys, resulting in unauthorized access or failed queries.
Knowledge Mining with Azure AI
Implementing Azure AI Search for Information Retrieval
Overlooking the need for a skillset when dealing with unstructured data like PDFs or images, which require AI enrichment to become searchable.
Knowledge Mining with Azure AI
Document Intelligence
Azure AI service for extracting data from documents.
Knowledge Mining with Azure AI
Key-Value Pairs
Identified field and its associated value (e.g., 'Total: $100').
Knowledge Mining with Azure AI
Prebuilt Model
Ready-to-use model for common document types (e.g., invoices).
Knowledge Mining with Azure AI
Custom Model
Model trained on user-provided data for specific document layouts.
Knowledge Mining with Azure AI
Layout Analysis
Identifies structural elements like paragraphs, titles, and tables.
Knowledge Mining with Azure AI
Extracting Insights with Document Intelligence
To remember Document Intelligence capabilities: 'DOC-INT' means 'Data Out Clearly, Instantly, Neatly, Through Intelligence.'
Knowledge Mining with Azure AI
Extracting Insights with Document Intelligence
For the AI-102 exam, remember that Document Intelligence (formerly Form Recognizer) excels at understanding document structure beyond simple OCR. Be prepared to differentiate between prebuilt models (invoices, receipts, ID documents) and custom models, and know when to use each. The 'Read' model is for general text extraction.
Knowledge Mining with Azure AI
Extracting Insights with Document Intelligence
Confusing Document Intelligence with simple OCR; DI understands context and structure.
Knowledge Mining with Azure AI
Extracting Insights with Document Intelligence
Trying to train a custom model for a document type that already has a highly accurate prebuilt model.
Knowledge Mining with Azure AI
Extracting Insights with Document Intelligence
Not providing enough diverse training examples for a custom model, leading to poor accuracy.
Knowledge Mining with Azure AI
Knowledge Mining
Extracting, enriching, and indexing insights from unstructured data.
Knowledge Mining with Azure AI
AI Enrichment
Applying AI services to extract entities, sentiment, and other metadata.
Knowledge Mining with Azure AI
Azure AI Document Intelligence
AI service for extracting data from documents via OCR and deep learning.
Knowledge Mining with Azure AI
Unstructured Data
Information without a predefined data model, like text or images.
Knowledge Mining with Azure AI
Building End-to-End Knowledge Mining Solutions
Imagine a 'KNOWLEDGE MINE' with a conveyor belt: Data goes IN, AI 'SKILLS' dig for gems, and a SEARCHLIGHT finds them in the INDEX.
Knowledge Mining with Azure AI
Building End-to-End Knowledge Mining Solutions
The exam frequently tests your understanding of the flow: Data Source -> Indexer -> Skillset (AI Enrichment) -> Index. Be prepared to identify which Azure AI service performs which specific enrichment task (e.g., Document Intelligence for forms, Language for sentiment).
Knowledge Mining with Azure AI
Building End-to-End Knowledge Mining Solutions
Confusing the role of an indexer with a skillset; the indexer pulls data, the skillset processes it.
Knowledge Mining with Azure AI
Building End-to-End Knowledge Mining Solutions
Underestimating the importance of data preparation before AI enrichment, as dirty data leads to poor insights.
Knowledge Mining with Azure AI
Building End-to-End Knowledge Mining Solutions
Forgetting that a user interface is often needed to consume the insights from the search index.
Knowledge Mining with Azure AI
Custom Scoring Profile
Defines how search results are ranked beyond default relevance.
Knowledge Mining with Azure AI
Text Analysis
Processing text for search, including tokenization and filtering.
Knowledge Mining with Azure AI
Synonyms
Mapping equivalent terms to improve search recall.
Knowledge Mining with Azure AI
Labeled Training Data
Documents with manually annotated fields for model training.
Knowledge Mining with Azure AI
Post-processing
Applying rules or logic to refine extracted data after AI processing.
Knowledge Mining with Azure AI
Incremental Indexing
Updating only changed parts of a search index.
Knowledge Mining with Azure AI
Custom Neural Model
A Document Intelligence model for highly variable document structures.
Knowledge Mining with Azure AI
Optimizing Search and Document Intelligence
To OPTIMIZE, remember: 'S.C.A.L.E.' - Schema, Custom Scoring, Accuracy, Labeled Data, Efficiency.
Knowledge Mining with Azure AI
Optimizing Search and Document Intelligence
The exam often tests your understanding of when to use specific Azure AI Search features like custom scoring profiles versus simple field weighting, and the difference between Document Intelligence model types (prebuilt, custom template, custom neural) for various document complexities. Keywords: 'relevance tuning', 'accuracy improvement', 'data preparation', 'indexing strategies'.
Knowledge Mining with Azure AI
Optimizing Search and Document Intelligence
Not using custom scoring profiles when default relevance is insufficient, leading to poor search results.
Knowledge Mining with Azure AI
Optimizing Search and Document Intelligence
Training Document Intelligence models with too few or unrepresentative samples, resulting in low accuracy.
Knowledge Mining with Azure AI
Optimizing Search and Document Intelligence
Indexing all document fields in Azure AI Search without considering their search relevance, increasing costs and slowing queries.
Knowledge Mining with Azure AI
Azure OpenAI Service
Microsoft's cloud service for OpenAI models with enterprise features.
Generative AI with Azure OpenAI Service
GPT-4
OpenAI's most advanced large language model.
Generative AI with Azure OpenAI Service
DALL-E
OpenAI's model for generating images from text prompts.
Generative AI with Azure OpenAI Service
Content Filtering
System to detect and filter harmful or inappropriate content.
Generative AI with Azure OpenAI Service
Private Endpoint
Network interface connecting Azure services privately to a VNet.
Generative AI with Azure OpenAI Service
Model Deployment
An instance of a specific AI model provisioned for use.
Generative AI with Azure OpenAI Service
REST API
Standard for web service communication, used for model access.
Generative AI with Azure OpenAI Service
Getting Started with Azure OpenAI Service
To remember the benefits of Azure OpenAI, think 'S.P.E.C.S.': Security, Privacy, Enterprise-grade, Compliance, Scalability.
Generative AI with Azure OpenAI Service
Getting Started with Azure OpenAI Service
Memorize that Azure OpenAI Service is a 'limited access' service, requiring an application process. Also, understand that data submitted to Azure OpenAI Service is NOT used to retrain OpenAI's foundational models, a key differentiator for enterprise use.
Generative AI with Azure OpenAI Service
Getting Started with Azure OpenAI Service
Assuming Azure OpenAI Service is automatically available without requesting access first.
Generative AI with Azure OpenAI Service
Getting Started with Azure OpenAI Service
Confusing Azure OpenAI Service with OpenAI's public API, especially regarding data privacy and security features.
Generative AI with Azure OpenAI Service
Getting Started with Azure OpenAI Service
Forgetting to deploy a specific model after creating an Azure OpenAI resource, leading to API errors.
Generative AI with Azure OpenAI Service
Zero-shot Prompting
Asking an AI model to perform a task without any examples.
Generative AI with Azure OpenAI Service
Few-shot Prompting
Providing an AI model with a few input-output examples.
Generative AI with Azure OpenAI Service
Chain-of-Thought (CoT)
Prompting technique to make AI explain its reasoning steps.
Generative AI with Azure OpenAI Service
Temperature
A parameter controlling the randomness or creativity of AI output.
Generative AI with Azure OpenAI Service
Top_p
A parameter controlling the diversity of AI output by token probability.
Generative AI with Azure OpenAI Service
Context Window
The maximum amount of text (tokens) an LLM can process at once.
Generative AI with Azure OpenAI Service
Mastering Prompt Engineering Techniques
To remember prompt components, think of 'I C I O E': Instructions, Context, Input Data, Output Format, Examples. It's like telling a robot exactly what to do, what to know, what to use, how to give it back, and showing it how to do it!
Generative AI with Azure OpenAI Service
Mastering Prompt Engineering Techniques
The AI-102 exam frequently tests understanding of prompt components and techniques. Memorize the difference between zero-shot, few-shot, and Chain-of-Thought prompting, and recognize scenarios where each would be most effective. Keywords like 'context,' 'examples,' 'instructions,' and 'output format' are crucial.
Generative AI with Azure OpenAI Service
Mastering Prompt Engineering Techniques
Providing overly vague or ambiguous instructions, leading to generic or irrelevant outputs.
Generative AI with Azure OpenAI Service
Mastering Prompt Engineering Techniques
Not iterating on prompts; expecting perfect output from the first attempt without refinement.
Generative AI with Azure OpenAI Service
Mastering Prompt Engineering Techniques
Ignoring the model's limitations or context window, resulting in truncated or incomplete responses.
Generative AI with Azure OpenAI Service
Deployment
An instance of an Azure OpenAI model made available for use.
Generative AI with Azure OpenAI Service
Retrieval-Augmented Generation (RAG)
A pattern combining LLMs with external data retrieval.
Generative AI with Azure OpenAI Service
Rate Limits
Restrictions on the number of API requests or tokens per minute.
Generative AI with Azure OpenAI Service
Token Usage
The measure of input and output text processed by an LLM, used for billing.
Generative AI with Azure OpenAI Service
Implementing Generative AI Solutions
To 'Implement' a solution, remember 'DREAM': Deploy, Retrieve, Engineer, Authenticate, Monitor.
Generative AI with Azure OpenAI Service
Implementing Generative AI Solutions
The exam frequently asks about the purpose and benefits of Retrieval-Augmented Generation (RAG) and the steps for deploying and consuming models. Look for keywords like 'grounding,' 'external data,' and 'factual accuracy' when RAG is discussed.
Generative AI with Azure OpenAI Service
Implementing Generative AI Solutions
Not securing API keys properly, leading to unauthorized access and potential cost overruns.
Generative AI with Azure OpenAI Service
Implementing Generative AI Solutions
Ignoring rate limits, causing applications to fail due to too many requests.
Generative AI with Azure OpenAI Service
Implementing Generative AI Solutions
Failing to implement RAG when factual accuracy with external, dynamic data is critical, leading to hallucinations.
Generative AI with Azure OpenAI Service
Responsible AI
Ethical and safe development and deployment of AI.
Generative AI with Azure OpenAI Service
Bias Amplification
AI increasing existing biases from training data.
Generative AI with Azure OpenAI Service
Abuse Monitoring
Tracking AI usage to detect policy violations.
Generative AI with Azure OpenAI Service
Human-in-the-Loop
Human oversight and intervention in AI processes.
Generative AI with Azure OpenAI Service
Transparency
Understanding how AI systems make decisions.
Generative AI with Azure OpenAI Service
Fairness
AI treating all user groups equitably without bias.
Generative AI with Azure OpenAI Service
Ensuring Responsible AI Practices in Generative AI
Think 'FRIPS-TA': Fairness, Reliability, Inclusiveness, Privacy, Security, Transparency, Accountability. These are the pillars of Responsible AI!
Generative AI with Azure OpenAI Service
Ensuring Responsible AI Practices in Generative AI
The exam often tests your understanding of Azure OpenAI's specific content filtering categories (hate, sexual, self-harm, violence) and that these filters are always active and configurable, not disable-able. Memorize the core Responsible AI principles.
Generative AI with Azure OpenAI Service
Ensuring Responsible AI Practices in Generative AI
Assuming Azure OpenAI's default content filters are sufficient for all applications; custom moderation is often needed.
Generative AI with Azure OpenAI Service
Ensuring Responsible AI Practices in Generative AI
Forgetting that content filters are always active and cannot be turned off, only configured for sensitivity.
Generative AI with Azure OpenAI Service
Ensuring Responsible AI Practices in Generative AI
Neglecting to inform users that content is AI-generated, leading to potential trust issues.
Generative AI with Azure OpenAI Service
Tree-of-Thought (ToT)
Prompting technique exploring multiple reasoning paths in parallel.
Generative AI with Azure OpenAI Service
Knowledge Base
Repository of structured or unstructured data for RAG systems.
Generative AI with Azure OpenAI Service
Vector Database
Database optimized for storing and querying vector embeddings.
Generative AI with Azure OpenAI Service
Advanced Prompt Engineering and Model Customization
RAG for Facts, Fine-tune for Style! RAG gets the 'F'actual data, Fine-tuning gives it 'S'tyle and consistency.
Generative AI with Azure OpenAI Service
Advanced Prompt Engineering and Model Customization
The exam often tests your understanding of when to use RAG versus fine-tuning. Remember, RAG is for up-to-date, factual, or proprietary data, while fine-tuning is for specific style, tone, or task consistency when you have a good dataset.
Generative AI with Azure OpenAI Service
Advanced Prompt Engineering and Model Customization
Trying to use RAG for style or tone adjustments instead of fine-tuning.
Generative AI with Azure OpenAI Service
Advanced Prompt Engineering and Model Customization
Attempting to fine-tune without a sufficiently large and high-quality dataset.
Generative AI with Azure OpenAI Service
Advanced Prompt Engineering and Model Customization
Overlooking Chain-of-Thought for complex reasoning, leading to less accurate model outputs.
Generative AI with Azure OpenAI Service