AI-102 Study Plan
This study plan is a structured, execution-oriented roadmap designed specifically for candidates preparing for the AI-102: Designing and Implementing a Microsoft Azure AI Solution exam.
Unlike generic reading lists or loosely organized schedules, this plan is built around how people actually learn and retain complex technical knowledge, and how the AI-102 exam actually evaluates understanding. Its primary purpose is not only to help you pass the exam, but to ensure that you develop correct architectural judgment, which is the core competency tested by AI-102.
Week 1: Plan and Manage an Azure AI Solution
Weekly Focus
This week builds the architectural foundation for the entire AI-102 exam.
The goal is not memorization, but correct service selection, deployment reasoning, and lifecycle management.
Everything you learn later (OpenAI, Agents, Vision, NLP, Search) depends on this week.
Weekly Learning Objectives
By the end of Week 1, you must be able to:
Explain the purpose and boundaries of each major Azure AI service in plain language.
Select the correct Azure AI service for a given business scenario and justify the choice.
Decide how an AI solution should be deployed (API, container, or edge) and explain why.
Describe how an Azure AI solution is monitored, scaled, and cost-controlled in production.
Explain Responsible AI requirements and compliance considerations at an exam-appropriate level.
If you cannot do all five, you are not ready to move on.
Learning Method Applied This Week
Daily structure:
4 to 6 Pomodoro sessions per day
Each Pomodoro is 25 minutes
Every day includes:
New content learning
Active processing (rewriting, comparing, explaining)
Spaced review based on the forgetting curve
No passive reading is allowed.
Day 1: Azure AI Landscape and Mental Model
Daily Objective
Build a correct mental model of what Azure AI is and what it is not.
Pomodoro 1: What Azure AI Is
Task:
Read the section explaining what Azure AI is.
Rewrite the concept in your own words focusing on:
Why Azure provides prebuilt AI services
Why most solutions do not require custom model training
Write a short paragraph answering:
- What problems Azure AI is designed to solve for enterprises
Expected outcome:
- You can explain Azure AI to a non-AI engineer in under two minutes.
Pomodoro 2: Azure Cognitive Services Overview
Task:
Study the four categories of Cognitive Services:
Vision
Speech
Language
Decision
For each category, write:
Typical input
Typical output
One real-world business use case
Do not write APIs or SDK details.
Expected outcome:
- You can recognize which Cognitive Service category fits a described task.
Pomodoro 3: Azure OpenAI Service Positioning
Task:
Study Azure OpenAI Service at a conceptual level.
Write a comparison table:
Azure Cognitive Services vs Azure OpenAI
Deterministic vs generative output
Cost and risk differences
Write a paragraph answering:
- Why enterprises cannot treat OpenAI like a normal API
Expected outcome:
- You understand why Azure OpenAI requires stricter governance.
Pomodoro 4: Same-Day Review
Task:
Without looking at notes, write a one-page explanation answering:
What Azure AI services exist
When to use Cognitive Services
When to use Azure OpenAI
Then check and correct gaps.
Forgetting curve action:
- This creates the first memory trace.
Day 2: Service Selection and Architectural Decision-Making
Daily Objective
Learn how the exam tests service selection through scenarios.
Pomodoro 1: Service Selection Criteria
Task:
Study the selection criteria:
Performance
Scalability
Cost
Security
Compliance
For each criterion, write:
- One example where it becomes the dominant factor
Expected outcome:
- You understand why the “best” service depends on constraints.
Pomodoro 2: Scenario-Based Service Choice
Task:
Take three example scenarios:
E-commerce
Healthcare
Enterprise knowledge base
For each scenario:
Choose services
Explain why alternatives are weaker choices
Expected outcome:
- You can justify a choice, not just name a service.
Pomodoro 3: Anti-Pattern Analysis
Task:
Write three incorrect architectural decisions, such as:
Using OpenAI without grounding
Using Azure ML instead of Cognitive Services
Explain why each decision is wrong.
Expected outcome:
- You recognize common exam traps.
Pomodoro 4: Review Day 1 Content
Task:
Re-explain Day 1 concepts in exam-style language.
Focus on service boundaries and terminology.
Forgetting curve action:
- T+1 reinforcement of Day 1.
Day 3: Deployment Models for Azure AI Solutions
Daily Objective
Understand how AI solutions are deployed and why deployment choice matters.
Pomodoro 1: API-Based Deployment
Task:
Study API-based deployment.
Write:
When API deployment is ideal
When it becomes a limitation
Provide one business example.
Expected outcome:
- You know why APIs are the default but not universal.
Pomodoro 2: Containerized Deployment
Task:
Study container deployment using AKS or ACI.
Write:
What control containers provide
Why regulated industries prefer this model
Expected outcome:
- You understand containers as a governance and control decision.
Pomodoro 3: Edge Deployment
Task:
Study edge AI deployment.
Write:
Why latency and privacy drive edge AI
Why edge AI is not suitable for all workloads
Expected outcome:
- You can explain edge AI without exaggeration.
Pomodoro 4: Review Day 2 Content
Task:
Review service selection decisions from Day 2.
Add deployment reasoning to each scenario.
Forgetting curve action:
- T+1 reinforcement of Day 2.
Day 4: Monitoring, Scaling, and Cost Management
Daily Objective
Understand how AI solutions are operated after deployment.
Pomodoro 1: Monitoring with Azure Monitor
Task:
Study Azure Monitor concepts.
Write:
What metrics matter for AI services
Why monitoring is critical for reliability
Pomodoro 2: Application Insights
Task:
Study Application Insights.
Write:
How it differs from Azure Monitor
What problems it helps diagnose
Pomodoro 3: Scaling and Cost Control
Task:
Study autoscaling and cost optimization.
Write:
Why AI cost can grow unexpectedly
How batching and caching reduce cost
Pomodoro 4: Review Day 1 and Day 3
Task:
- Review deployment models and service choices together.
Forgetting curve action:
- T+3 reinforcement of Day 1.
Day 5: Responsible AI and Compliance
Daily Objective
Understand ethical, legal, and governance expectations.
Pomodoro 1: Responsible AI Principles
Task:
Study the six Microsoft Responsible AI principles.
Write one concrete failure example per principle.
Pomodoro 2: Compliance Concepts
Task:
Study GDPR and HIPAA at a conceptual level.
Write:
What the system must do
What the system must never do
Pomodoro 3: Human-in-the-Loop
Task:
Study human oversight patterns.
Write:
When human review is mandatory
Why automation alone is risky
Pomodoro 4: Review Day 2 and Day 4
Task:
- Re-evaluate earlier scenarios with compliance in mind.
Forgetting curve action:
- T+3 reinforcement of Day 2.
Day 6: Integration and System Thinking
Daily Objective
Integrate everything into a single architectural mindset.
Pomodoro 1–2: End-to-End Architecture Exercise
Task:
Design a full Azure AI solution for:
- A document-processing enterprise system
Include:
Service selection
Deployment
Monitoring
Compliance
Pomodoro 3: Exam-Style Explanation
Task:
- Write explanations as if answering a case study question.
Pomodoro 4: Review Day 3 and Day 5
Forgetting curve action:
- T+3 reinforcement.
Day 7: Weekly Consolidation and Long-Term Retention
Daily Objective
Lock knowledge into long-term memory.
Pomodoro 1: Closed-Book Recall
Task:
- Write everything you remember about Week 1 without notes.
Pomodoro 2: Gap Analysis
Task:
Compare with notes.
Identify weak areas.
Pomodoro 3: Exam Language Refinement
Task:
- Rewrite explanations using precise AI-102 terminology.
Pomodoro 4: Preparation for Week 2
Task:
- Write three questions you expect to answer with Generative AI.
Forgetting curve action:
- T+7 reinforcement of Day 1.
Week 2: Implement Generative AI Solutions
Weekly Focus
This week focuses on generative AI in enterprise systems, specifically how Azure OpenAI is used safely, reliably, and correctly.
The exam does not test creativity.
It tests engineering judgment.
Weekly Learning Objectives
By the end of Week 2, you must be able to:
Explain how Azure OpenAI differs from public OpenAI services.
Decide when generative AI is appropriate and when it is not.
Design a Retrieval-Augmented Generation (RAG) solution at a conceptual level.
Explain why grounding, security, and cost control are mandatory in enterprise AI.
Identify common generative AI failure modes and how Azure mitigates them.
You should be able to explain these clearly without referencing APIs or code.
Learning Method Applied This Week
This week emphasizes:
Conceptual precision
Scenario reasoning
Error analysis
Daily structure:
4 to 6 Pomodoro sessions
Every day includes:
New concept learning
Scenario-based reasoning
Spaced review from Week 1
Day 1: Azure OpenAI Service Fundamentals
Daily Objective
Understand what Azure OpenAI is designed to do in enterprise environments.
Pomodoro 1: Azure OpenAI Purpose and Scope
Task:
Study the conceptual description of Azure OpenAI Service.
Write a short explanation answering:
Why Microsoft offers OpenAI models through Azure
Why enterprises prefer Azure OpenAI over public endpoints
Expected outcome:
- You can explain Azure OpenAI in business terms, not technical terms.
Pomodoro 2: Model Capabilities and Limitations
Task:
Study the types of generative tasks supported:
Text generation
Summarization
Question answering
Write a list of tasks that generative AI should not be trusted to do alone.
Expected outcome:
- You understand that generative AI is probabilistic, not authoritative.
Pomodoro 3: Deterministic vs Generative Output
Task:
Compare:
Cognitive Services output
Azure OpenAI output
Write examples where determinism is required.
Expected outcome:
- You can justify why generative AI is risky without controls.
Pomodoro 4: Same-Day Review
Task:
Write a half-page explanation:
When Azure OpenAI is appropriate
When it is not appropriate
Forgetting curve action:
- First memory encoding.
Day 2: Prompting as an Engineering Tool
Daily Objective
Understand prompting as system behavior control, not creative writing.
Pomodoro 1: Role of Prompts in Enterprise AI
Task:
Study how prompts guide behavior.
Write:
What prompts can control
What prompts cannot guarantee
Expected outcome:
- You treat prompts as guardrails, not solutions.
Pomodoro 2: System Instructions and Constraints
Task:
Study system-level instructions.
Write examples of:
Safety constraints
Behavioral constraints
Explain why these are critical for compliance.
Expected outcome:
- You understand prompts as policy, not conversation.
Pomodoro 3: Prompt Failure Modes
Task:
Write examples of:
Hallucination
Overconfidence
Ambiguous responses
Explain why these failures occur.
Expected outcome:
- You can recognize unsafe generative behavior.
Pomodoro 4: Review Day 1
Task:
- Re-explain Azure OpenAI fundamentals using exam-style language.
Forgetting curve action:
- T+1 reinforcement of Day 1.
Day 3: Retrieval-Augmented Generation (RAG) Concepts
Daily Objective
Understand why RAG is essential for enterprise AI.
Pomodoro 1: Why Pure Generative AI Fails in Enterprises
Task:
Study the limitations of standalone LLMs.
Write a paragraph explaining:
- Why models cannot be trusted as knowledge sources
Expected outcome:
- You understand the need for grounding.
Pomodoro 2: RAG Architecture Overview
Task:
Study the high-level RAG flow:
User query
Retrieval
Generation
Draw a simple conceptual diagram in words.
Expected outcome:
- You can explain RAG without technical jargon.
Pomodoro 3: Role of Azure AI Search in RAG
Task:
Study how search indexes provide grounding.
Write:
- Why AI Search is preferred over dumping documents into prompts
Expected outcome:
- You understand retrieval as a reliability mechanism.
Pomodoro 4: Review Day 2
Forgetting curve action:
- T+1 reinforcement of Day 2.
Day 4: Grounding, Safety, and Data Protection
Daily Objective
Understand how Azure controls generative AI risk.
Pomodoro 1: Grounding Strategies
Task:
Study grounding concepts:
Retrieved documents
Citations
Write why grounding reduces hallucination but does not eliminate it.
Pomodoro 2: Data Privacy and Training Boundaries
Task:
Study enterprise data protection concepts.
Write:
What data is not used for model training
Why this matters legally and ethically
Expected outcome:
- You can answer exam questions about data usage confidently.
Pomodoro 3: Content Filtering and Safety
Task:
Study content filtering at a conceptual level.
Write examples of:
Harmful content
Regulatory risks
Expected outcome:
- You understand safety as a system requirement.
Pomodoro 4: Review Day 1 and Day 3
Forgetting curve action:
- T+3 reinforcement of Day 1.
Day 5: Cost, Performance, and Reliability
Daily Objective
Understand why generative AI requires strict operational controls.
Pomodoro 1: Token-Based Cost Model
Task:
Study how generative AI cost is calculated conceptually.
Write:
Why costs scale unpredictably
Why cost monitoring is essential
Pomodoro 2: Latency and User Experience
Task:
Study latency considerations.
Write:
Why generative AI is slower than traditional APIs
How this affects architecture decisions
Pomodoro 3: Reliability and Fallback Strategies
Task:
Write examples of:
Fallback responses
Graceful degradation
Explain why systems must handle model failure.
Pomodoro 4: Review Day 3 and Day 4
Forgetting curve action:
- T+3 reinforcement of Day 3.
Day 6: Scenario Integration and Decision-Making
Daily Objective
Apply generative AI concepts to real exam-style scenarios.
Pomodoro 1–2: Scenario Design Exercise
Task:
Design a generative AI solution for:
- An internal employee Q&A system
Include:
Whether RAG is needed
How data is protected
How cost is controlled
Pomodoro 3: Decision Justification
Task:
- Write exam-style justifications for each architectural choice.
Pomodoro 4: Review Day 5
Forgetting curve action:
- T+3 reinforcement.
Day 7: Weekly Consolidation and Retention
Daily Objective
Convert knowledge into long-term memory.
Pomodoro 1: Closed-Book Recall
Task:
Write everything you remember about:
Azure OpenAI
RAG
Safety
Cost
Pomodoro 2: Gap Identification
Task:
- Identify weak areas and restudy selectively.
Pomodoro 3: Exam Language Refinement
Task:
- Rewrite explanations using precise AI-102 terminology.
Pomodoro 4: Preparation for Week 3
Task:
- Write three questions about multi-step or action-based AI workflows.
Forgetting curve action:
- T+7 reinforcement of Day 1.
Week 3: Implement an Agentic Solution
Weekly Focus
This week focuses on agentic AI systems, which are tested in AI-102 as goal-driven, multi-step, action-capable solutions.
The exam does not expect you to build agents from scratch.
It expects you to recognize when agents are required, how they are structured, and how risk is controlled.
Weekly Learning Objectives
By the end of Week 3, you must be able to:
Clearly distinguish between:
A chatbot
A RAG-based system
An agentic solution
Explain why tool calling is essential for reliable agent behavior.
Describe the core components of an agentic system:
Model
Tools
Memory
Orchestration
Identify risks introduced by agents and explain how they are mitigated.
Decide when an agent is justified and when it is unnecessary or harmful.
You should be able to explain these decisions in exam-style reasoning.
Learning Method Applied This Week
This week emphasizes:
Systems thinking
Step-by-step reasoning
Failure analysis
Daily structure:
4 to 6 Pomodoro sessions
Every day includes:
Conceptual learning
Decomposition of agent behavior
Spaced review of generative AI concepts from Week 2
Day 1: What “Agentic” Means in AI-102
Daily Objective
Understand what an agentic solution is in the exam context, and what it is not.
Pomodoro 1: Defining an Agentic Solution
Task:
Study the definition of an agentic solution in Azure AI.
Write a clear explanation answering:
How an agent differs from a chatbot
How an agent differs from a RAG system
Expected outcome:
- You can define an agent in one precise paragraph.
Pomodoro 2: Goal-Driven vs Response-Driven Systems
Task:
Compare:
Single-response systems
Goal-driven multi-step systems
Write examples where a single response is insufficient.
Expected outcome:
- You understand why agents exist at all.
Pomodoro 3: Exam Framing of Agents
Task:
Study how AI-102 frames agents as workflow orchestrators.
Write:
What the exam expects you to know
What the exam does not test
Expected outcome:
- You avoid overengineering your understanding.
Pomodoro 4: Same-Day Review
Task:
Write a short explanation:
When an agent is the correct architectural choice
When it is a mistake
Forgetting curve action:
- Initial memory encoding.
Day 2: Tools and Function Calling
Daily Objective
Understand tools as the foundation of reliable agent behavior.
Pomodoro 1: Purpose of Tool Calling
Task:
Study why agents must call tools.
Write:
Why text-only reasoning is unreliable
Why tools provide authoritative results
Expected outcome:
- You understand tools as trust anchors.
Pomodoro 2: Types of Tools
Task:
Study different tool categories:
Retrieval tools
Action tools
Computation tools
For each category, write one enterprise example.
Expected outcome:
- You can classify tools correctly in scenarios.
Pomodoro 3: Tool Schema and Reliability
Task:
Study why structured inputs and outputs matter.
Write:
How schemas reduce ambiguity
Why validation is mandatory
Expected outcome:
- You understand tools as contracts, not helpers.
Pomodoro 4: Review Day 1
Forgetting curve action:
- T+1 reinforcement of agent definition.
Day 3: Memory and State Management
Daily Objective
Understand how agents maintain continuity and context.
Pomodoro 1: Short-Term State
Task:
Study session-level memory.
Write:
What information must persist during a task
What should not persist
Expected outcome:
- You understand session scoping.
Pomodoro 2: Long-Term Memory
Task:
Study long-term memory concepts.
Write:
When storing user preferences is useful
Why long-term memory introduces privacy risk
Expected outcome:
- You can reason about retention decisions.
Pomodoro 3: Auditing and Traceability
Task:
Study why agent actions must be traceable.
Write:
What should be logged
What must not be logged
Expected outcome:
- You understand compliance-driven logging.
Pomodoro 4: Review Day 2
Forgetting curve action:
- T+1 reinforcement of tool concepts.
Day 4: Orchestration and Control Flow
Daily Objective
Understand how agent workflows are executed safely.
Pomodoro 1: Agent Execution Loop
Task:
Study the standard agent loop:
Interpret
Plan
Act
Observe
Iterate
Rewrite this loop in your own words.
Expected outcome:
- You understand the agent lifecycle.
Pomodoro 2: Error Handling and Retries
Task:
Study failure scenarios:
Tool timeout
Partial results
Invalid responses
Write how an agent should respond safely.
Expected outcome:
- You understand failure as a normal condition.
Pomodoro 3: Preventing Uncontrolled Execution
Task:
Study guardrails and stopping conditions.
Write:
Why agents must not loop indefinitely
How termination is enforced
Expected outcome:
- You understand risk containment.
Pomodoro 4: Review Day 1 and Day 3
Forgetting curve action:
- T+3 reinforcement of core agent concepts.
Day 5: Safety, Security, and Governance for Agents
Daily Objective
Understand why agents amplify risk and how Azure mitigates it.
Pomodoro 1: Prompt Injection and Tool Abuse
Task:
Study how agents can be manipulated.
Write:
Why retrieved content is untrusted
How tool access must be restricted
Expected outcome:
- You understand agents as attack surfaces.
Pomodoro 2: Access Control and Permissions
Task:
Study identity-based access for tools.
Write:
Why least privilege is essential
Why agents must respect user permissions
Expected outcome:
- You can reason about security boundaries.
Pomodoro 3: Human Oversight for Agents
Task:
Study approval gates.
Write:
Which actions require human approval
Why full automation is unsafe
Expected outcome:
- You understand accountability requirements.
Pomodoro 4: Review Day 2 and Day 4
Forgetting curve action:
- T+3 reinforcement.
Day 6: Scenario-Based Agent Design
Daily Objective
Apply agentic concepts to realistic exam scenarios.
Pomodoro 1–2: Scenario Exercise
Task:
Design an agent for:
- IT support ticket handling
Decide:
Which tools are required
Which steps are automated
Where human review is needed
Pomodoro 3: Architectural Justification
Task:
Write exam-style justifications for:
Why an agent is used
Why alternatives are insufficient
Pomodoro 4: Review Day 5
Forgetting curve action:
- T+3 reinforcement.
Day 7: Weekly Consolidation and Retention
Daily Objective
Convert understanding into long-term memory.
Pomodoro 1: Closed-Book Recall
Task:
- Write everything you remember about agentic systems.
Pomodoro 2: Weakness Identification
Task:
- Identify unclear concepts and restudy selectively.
Pomodoro 3: Exam Language Refinement
Task:
- Rewrite explanations using precise AI-102 terminology.
Pomodoro 4: Preparation for Week 4
Task:
- Write three questions related to image and document processing.
Forgetting curve action:
- T+7 reinforcement of Day 1.
Week 4: Implement Computer Vision Solutions
Weekly Focus
This week focuses on computer vision services in Azure, with an emphasis on service selection, capability boundaries, and compliance-aware design.
The exam does not test image-processing theory.
It tests whether you can choose the correct vision service and apply it appropriately.
Weekly Learning Objectives
By the end of Week 4, you must be able to:
Distinguish between Azure Computer Vision, Custom Vision, and Document Intelligence.
Select the correct vision service for image analysis, OCR, and document processing scenarios.
Explain when prebuilt models are sufficient and when custom models are required.
Identify compliance and ethical risks related to visual data, especially face recognition.
Explain how vision solutions are integrated, deployed, and scaled in enterprise systems.
You should be able to justify all choices in exam-style scenario questions.
Learning Method Applied This Week
This week emphasizes:
Input–output thinking
Service boundary clarity
Scenario-driven differentiation
Daily structure:
4 to 6 Pomodoro sessions
Every day includes:
New vision concepts
Comparative reasoning between services
Spaced review from Weeks 2 and 3
Day 1: Azure Computer Vision Fundamentals
Daily Objective
Understand what Azure Computer Vision can and cannot do.
Pomodoro 1: Purpose and Scope of Azure Computer Vision
Task:
Study the core purpose of Azure Computer Vision.
Write a clear explanation covering:
What types of images it analyzes
What kinds of insights it returns
Avoid implementation details.
Expected outcome:
- You can explain Computer Vision as a capability, not an API.
Pomodoro 2: Core Capabilities Overview
Task:
Study the following features conceptually:
Image analysis
Object detection
Image tagging
Image captioning
For each feature, write:
Typical input
Typical output
One business use case
Expected outcome:
- You can match features to use cases accurately.
Pomodoro 3: Strengths and Limitations
Task:
Write:
Three problems Azure Computer Vision solves well
Three problems it should not be used for
Explain why limitations exist.
Expected outcome:
- You avoid overusing vision services.
Pomodoro 4: Same-Day Review
Task:
- Write a short summary explaining when Azure Computer Vision is the correct choice.
Forgetting curve action:
- Initial memory encoding.
Day 2: Optical Character Recognition (OCR)
Daily Objective
Understand text extraction from images and documents.
Pomodoro 1: OCR Concepts and Use Cases
Task:
Study what OCR does conceptually.
Write examples of:
Printed text extraction
Handwritten text extraction
Explain why OCR is critical for document automation.
Expected outcome:
- You understand OCR as a transformation step.
Pomodoro 2: OCR vs Image Analysis
Task:
Compare OCR with general image analysis.
Write:
Why OCR is not just “reading text from images”
Why structured output matters
Expected outcome:
- You avoid confusing OCR with tagging.
Pomodoro 3: OCR Limitations and Errors
Task:
Write examples of:
Low-quality image issues
Language and layout challenges
Explain how these affect system design.
Expected outcome:
- You understand OCR reliability constraints.
Pomodoro 4: Review Day 1
Forgetting curve action:
- T+1 reinforcement of vision fundamentals.
Day 3: Document Intelligence (Form Recognizer)
Daily Objective
Understand structured document extraction.
Pomodoro 1: Purpose of Document Intelligence
Task:
Study what Document Intelligence is designed to do.
Write:
Why it exists separately from OCR
What “structured extraction” means
Expected outcome:
- You understand its role in enterprise workflows.
Pomodoro 2: Prebuilt vs Custom Models
Task:
Study prebuilt models (invoices, receipts, IDs).
Compare with custom document models.
Write when each is appropriate.
Expected outcome:
- You can justify model choice.
Pomodoro 3: Business Process Integration
Task:
Write how Document Intelligence fits into:
Accounts payable
Contract processing
Compliance workflows
Expected outcome:
- You see document AI as part of a pipeline.
Pomodoro 4: Review Day 2
Forgetting curve action:
- T+1 reinforcement of OCR concepts.
Day 4: Custom Vision and Model Customization
Daily Objective
Understand when and why custom vision models are required.
Pomodoro 1: Why Custom Vision Exists
Task:
Study limitations of prebuilt vision models.
Write scenarios where domain-specific recognition is required.
Expected outcome:
- You understand the justification for customization.
Pomodoro 2: Training and Data Considerations
Task:
Study training data requirements at a conceptual level.
Write:
Why labeling quality matters
Why dataset bias affects results
Expected outcome:
- You understand training risk, not training mechanics.
Pomodoro 3: Deployment and Maintenance Tradeoffs
Task:
Write:
Why custom models cost more to maintain
When maintenance effort is justified
Expected outcome:
- You understand operational impact.
Pomodoro 4: Review Day 1 and Day 3
Forgetting curve action:
- T+3 reinforcement of core vision services.
Day 5: Face Recognition and Compliance
Daily Objective
Understand sensitive vision capabilities and governance requirements.
Pomodoro 1: Face Detection vs Face Recognition
Task:
Study the difference conceptually.
Write:
What face detection does
What face recognition does
Explain why the distinction matters legally.
Expected outcome:
- You avoid compliance mistakes.
Pomodoro 2: Ethical and Legal Risks
Task:
Study privacy and bias concerns.
Write:
Why face recognition is high-risk
When it should be avoided
Expected outcome:
- You understand Responsible AI implications.
Pomodoro 3: Governance and Safeguards
Task:
Write:
Required controls for face-related systems
Role of human oversight
Expected outcome:
- You can answer compliance-based exam questions.
Pomodoro 4: Review Day 3 and Day 4
Forgetting curve action:
- T+3 reinforcement.
Day 6: Vision Solution Architecture
Daily Objective
Integrate vision services into complete systems.
Pomodoro 1–2: End-to-End Scenario
Task:
Design a vision-based solution for:
- Automated document processing
Decide:
Which vision service is used at each step
How results are stored and validated
Pomodoro 3: Architecture Justification
Task:
- Write exam-style justifications for each design decision.
Pomodoro 4: Review Day 5
Forgetting curve action:
- T+3 reinforcement.
Day 7: Weekly Consolidation and Retention
Daily Objective
Lock vision knowledge into long-term memory.
Pomodoro 1: Closed-Book Recall
Task:
- Write everything you remember about computer vision services.
Pomodoro 2: Weak Area Identification
Task:
- Identify confusion points and restudy selectively.
Pomodoro 3: Exam Language Refinement
Task:
- Rewrite explanations using precise AI-102 terminology.
Pomodoro 4: Preparation for Week 5
Task:
- Write three questions related to text, language, or speech processing.
Forgetting curve action:
- T+7 reinforcement of Day 1.
Week 5: Implement Natural Language Processing Solutions
Weekly Focus
This week focuses on language understanding and speech processing using Azure AI services.
The exam does not test linguistic theory.
It tests whether you can select the correct language service, understand capability boundaries, and integrate NLP safely into enterprise systems.
Weekly Learning Objectives
By the end of Week 5, you must be able to:
Distinguish between Azure AI Language services and Azure OpenAI for text-based tasks.
Select the correct NLP capability for sentiment analysis, entity extraction, classification, and summarization.
Understand custom language models and when they are justified.
Explain speech-to-text and text-to-speech use cases and constraints.
Design NLP solutions that are secure, scalable, and appropriate for enterprise use.
You should be able to justify every service choice using exam-style reasoning.
Learning Method Applied This Week
This week emphasizes:
Intent-to-service mapping
Boundary awareness between deterministic NLP and generative AI
Scenario-driven differentiation
Daily structure:
4 to 6 Pomodoro sessions
Every day includes:
New NLP concepts
Service comparison exercises
Spaced review from Week 4 (Vision) and earlier weeks
Day 1: Azure AI Language Service Fundamentals
Daily Objective
Understand what Azure AI Language services are designed to do.
Pomodoro 1: Purpose and Scope of Azure AI Language
Task:
Study the role of Azure AI Language services.
Write a concise explanation covering:
What types of text it processes
What kinds of structured outputs it produces
Avoid references to APIs or SDKs.
Expected outcome:
- You can explain Azure AI Language as a business capability.
Pomodoro 2: Core Language Capabilities Overview
Task:
Study the following capabilities conceptually:
Language detection
Sentiment analysis
Key phrase extraction
Named entity recognition
For each capability, write:
Typical input
Typical output
One realistic business scenario
Expected outcome:
- You can map text-processing needs to the correct capability.
Pomodoro 3: Strengths and Limitations
Task:
Write:
Three problems Azure AI Language solves well
Three problems it should not be used for
Explain why these limitations exist.
Expected outcome:
- You avoid misusing deterministic NLP.
Pomodoro 4: Same-Day Review
Task:
- Write a summary explaining when Azure AI Language is preferred over Azure OpenAI.
Forgetting curve action:
- Initial memory encoding.
Day 2: Text Classification and Information Extraction
Daily Objective
Understand how text can be categorized and structured.
Pomodoro 1: Text Classification Concepts
Task:
Study text classification at a conceptual level.
Write:
What classification does
How it differs from sentiment analysis
Expected outcome:
- You understand intent categorization.
Pomodoro 2: Custom Classification Models
Task:
Study custom text classification.
Write:
When prebuilt models are insufficient
Why domain-specific labels matter
Expected outcome:
- You can justify customization decisions.
Pomodoro 3: Entity and Key Information Extraction
Task:
Study entity extraction use cases.
Write:
How entities support downstream automation
Why structured extraction is valuable
Expected outcome:
- You understand NLP as an automation enabler.
Pomodoro 4: Review Day 1
Forgetting curve action:
- T+1 reinforcement of language fundamentals.
Day 3: Azure AI Language vs Azure OpenAI
Daily Objective
Clearly separate deterministic NLP from generative AI.
Pomodoro 1: Deterministic vs Generative Text Processing
Task:
Compare:
Azure AI Language outputs
Azure OpenAI outputs
Write examples where determinism is required.
Expected outcome:
- You understand why OpenAI is not always appropriate.
Pomodoro 2: Service Selection Scenarios
Task:
Take three scenarios:
Customer feedback analysis
Legal document review
Internal knowledge summarization
Choose the correct service for each and explain why.
Expected outcome:
- You can make defensible service choices.
Pomodoro 3: Hybrid Approaches
Task:
Study scenarios combining Language services and OpenAI.
Write:
Why hybrid approaches improve reliability
Where boundaries must be enforced
Expected outcome:
- You understand layered architectures.
Pomodoro 4: Review Day 2
Forgetting curve action:
- T+1 reinforcement of classification concepts.
Day 4: Speech Services (Speech-to-Text and Text-to-Speech)
Daily Objective
Understand speech processing capabilities and constraints.
Pomodoro 1: Speech-to-Text Concepts
Task:
Study speech recognition at a conceptual level.
Write:
Typical inputs and outputs
Common enterprise use cases
Expected outcome:
- You understand speech as a data ingestion method.
Pomodoro 2: Text-to-Speech Concepts
Task:
Study speech synthesis.
Write:
Why voice output is used
Accessibility and UX considerations
Expected outcome:
- You understand speech as an interaction layer.
Pomodoro 3: Speech Limitations and Challenges
Task:
Write:
Accent and noise challenges
Latency considerations
Explain how these affect system design.
Expected outcome:
- You understand reliability constraints.
Pomodoro 4: Review Day 1 and Day 3
Forgetting curve action:
- T+3 reinforcement of core NLP concepts.
Day 5: Security, Compliance, and Privacy in NLP
Daily Objective
Understand governance requirements for text and speech data.
Pomodoro 1: Sensitive Text and PII
Task:
Study how NLP systems handle sensitive data.
Write:
What constitutes PII
Why text data is high-risk
Expected outcome:
- You understand privacy implications.
Pomodoro 2: Data Storage and Retention
Task:
Study retention considerations.
Write:
What data should be stored
What data should be discarded
Expected outcome:
- You understand compliance-driven design.
Pomodoro 3: Human Oversight and Auditing
Task:
Write:
When NLP outputs require review
How audits support accountability
Expected outcome:
- You understand operational governance.
Pomodoro 4: Review Day 3 and Day 4
Forgetting curve action:
- T+3 reinforcement.
Day 6: End-to-End NLP Solution Design
Daily Objective
Integrate language and speech into complete systems.
Pomodoro 1–2: Scenario-Based Design
Task:
Design an NLP solution for:
- Customer support ticket triage
Decide:
Which language capabilities are used
Whether generative AI is involved
How outputs are validated
Pomodoro 3: Architectural Justification
Task:
- Write exam-style explanations for each decision.
Pomodoro 4: Review Day 5
Forgetting curve action:
- T+3 reinforcement.
Day 7: Weekly Consolidation and Retention
Daily Objective
Convert NLP knowledge into long-term memory.
Pomodoro 1: Closed-Book Recall
Task:
- Write everything you remember about NLP and speech services.
Pomodoro 2: Weak Area Identification
Task:
- Identify unclear areas and restudy selectively.
Pomodoro 3: Exam Language Refinement
Task:
- Rewrite explanations using precise AI-102 terminology.
Pomodoro 4: Preparation for Week 6
Task:
- Write three questions related to enterprise search and knowledge mining.
Forgetting curve action:
- T+7 reinforcement of Day 1.
Week 6: Implement Knowledge Mining and Information Extraction Solutions, Plus Final Integration and Exam Readiness
Weekly Focus
This final week focuses on enterprise knowledge mining, where AI systems extract, enrich, index, and retrieve information from large volumes of documents.
This is where search, language, vision, and generative AI come together.
The exam heavily uses knowledge mining scenarios to test whether you can design end-to-end AI solutions, not isolated features.
Weekly Learning Objectives
By the end of Week 6, you must be able to:
Explain what knowledge mining is and why it is critical in enterprise AI systems.
Design a document ingestion and enrichment pipeline conceptually.
Explain how unstructured data becomes searchable and usable.
Decide when to combine search with NLP, vision, and generative AI.
Answer full AI-102 case-study questions with correct architectural reasoning.
You should be able to reason across all six AI-102 domains fluently.
Learning Method Applied This Week
This week emphasizes:
Integration of all prior knowledge
End-to-end system thinking
Exam-style synthesis
Daily structure:
4 to 6 Pomodoro sessions
Every day includes:
Knowledge mining concepts
Cross-domain integration
Spaced review from Weeks 1–5
Day 1: Knowledge Mining Fundamentals
Daily Objective
Understand what knowledge mining is and why enterprises need it.
Pomodoro 1: What Knowledge Mining Solves
Task:
Study the concept of knowledge mining.
Write a clear explanation answering:
What problem knowledge mining addresses
Why traditional databases are insufficient
Expected outcome:
- You understand knowledge mining as an information-access problem.
Pomodoro 2: Structured vs Unstructured Data
Task:
Study differences between structured and unstructured data.
Write:
Examples of each
Why most enterprise data is unstructured
Expected outcome:
- You understand why AI is required for enterprise search.
Pomodoro 3: Knowledge Mining Use Cases
Task:
Write explanations for:
Internal document search
Compliance and audit support
Customer support knowledge bases
Expected outcome:
- You can recognize knowledge mining scenarios instantly.
Pomodoro 4: Same-Day Review
Task:
- Write a summary explaining why knowledge mining is a core AI-102 topic.
Forgetting curve action:
- Initial memory encoding.
Day 2: Document Ingestion and Enrichment Pipelines
Daily Objective
Understand how raw documents become searchable knowledge.
Pomodoro 1: Document Ingestion Concepts
Task:
Study how documents enter a knowledge mining system.
Write:
Common document sources
Why ingestion must be automated
Expected outcome:
- You understand ingestion as a continuous process.
Pomodoro 2: Cognitive Enrichment
Task:
Study enrichment at a conceptual level.
Write:
What enrichment adds to raw text
Why enrichment improves retrieval quality
Expected outcome:
- You understand enrichment as value creation.
Pomodoro 3: Role of Vision and NLP in Enrichment
Task:
Write how:
OCR extracts text
NLP extracts entities and key phrases
Explain why multiple AI services are chained.
Expected outcome:
- You understand cross-service integration.
Pomodoro 4: Review Day 1
Forgetting curve action:
- T+1 reinforcement of fundamentals.
Day 3: Indexing, Search, and Retrieval
Daily Objective
Understand how enriched data becomes searchable.
Pomodoro 1: Search Index Concepts
Task:
Study what a search index represents.
Write:
Why indexes are not raw document stores
What fields matter for search
Expected outcome:
- You understand indexing as transformation, not storage.
Pomodoro 2: Query Types and Retrieval
Task:
Study:
Keyword search
Semantic search
Write when each is appropriate.
Expected outcome:
- You understand intent-based retrieval.
Pomodoro 3: Search Quality and Relevance
Task:
Write:
Why poor indexing causes poor answers
Why search quality impacts downstream AI
Expected outcome:
- You understand retrieval as a dependency.
Pomodoro 4: Review Day 2
Forgetting curve action:
- T+1 reinforcement of pipeline concepts.
Day 4: Knowledge Mining with Generative AI
Daily Objective
Understand how generative AI uses retrieved knowledge safely.
Pomodoro 1: Why Generative AI Needs Retrieval
Task:
Write:
Why generative models cannot be trusted as knowledge sources
How retrieval grounds responses
Expected outcome:
- You understand RAG as a reliability mechanism.
Pomodoro 2: Answer Generation from Retrieved Content
Task:
Study how retrieved data is summarized and reformulated.
Write:
Why citations matter
Why hallucination is still possible
Expected outcome:
- You understand limitations clearly.
Pomodoro 3: Failure Modes in Knowledge-Based Systems
Task:
Write examples of:
Missing documents
Outdated content
Incorrect enrichment
Explain mitigation strategies.
Expected outcome:
- You understand system-level risk.
Pomodoro 4: Review Day 1 and Day 3
Forgetting curve action:
- T+3 reinforcement of foundational knowledge.
Day 5: Security, Access Control, and Compliance
Daily Objective
Understand governance requirements in enterprise knowledge systems.
Pomodoro 1: Access Control and Permissions
Task:
Study document-level security concepts.
Write:
Why search must respect user permissions
Risks of overexposed knowledge
Expected outcome:
- You understand security as mandatory, not optional.
Pomodoro 2: Sensitive Information Handling
Task:
Study PII and sensitive document handling.
Write:
What should be indexed
What should be excluded or redacted
Expected outcome:
- You understand compliance-driven filtering.
Pomodoro 3: Auditing and Monitoring
Task:
Write:
Why knowledge systems must be auditable
What logs are required
Expected outcome:
- You understand operational accountability.
Pomodoro 4: Review Day 3 and Day 4
Forgetting curve action:
- T+3 reinforcement.
Day 6: Full AI-102 Case-Study Integration
Daily Objective
Integrate all AI-102 domains into a single solution mindset.
Pomodoro 1–2: End-to-End Case Study
Task:
Design a complete AI solution for:
- An enterprise document intelligence and Q&A system
Include:
Service selection
Ingestion and enrichment
Search and retrieval
Generative AI usage
Security and compliance
Pomodoro 3: Exam-Style Justification
Task:
Write answers as if responding to a case-study question:
Why each service is used
Why alternatives are rejected
Pomodoro 4: Review Day 5
Forgetting curve action:
- T+3 reinforcement.
Day 7: Final Consolidation and Exam Readiness
Daily Objective
Lock all AI-102 knowledge into long-term memory and exam readiness.
Pomodoro 1: Closed-Book Full Recall
Task:
Write everything you remember about:
- All six AI-102 domains
Pomodoro 2: Weak Area Diagnosis
Task:
Identify remaining weak topics.
Create a focused revision list.
Pomodoro 3: Exam Language Polishing
Task:
- Rewrite explanations using concise, precise exam terminology.
Pomodoro 4: Final Mental Rehearsal
Task:
Mentally walk through:
- A full AI-102 case-study from start to finish
Forgetting curve action:
- T+7 reinforcement of Week 1.