Developing Agentic & Generative AI Applications on Microsoft Azure

1. Introduction to Generative AI on Azure
   • Overview of Generative AI and LLMs
   • Azure OpenAI Service architecture
   • Foundation models and deployment models
   • Tokenization, context windows, pricing considerations
   • Responsible AI fundamentals
2. Deploy and Manage Azure OpenAI Models
   • Provision Azure OpenAI resource
   • Model deployment lifecycle
   • Role-Based Access Control (RBAC)
   • Secure network access & private endpoints
   • SDK vs REST integration
3. Prompt Engineering & Prompt Flow
   • Structured prompting techniques
   • Few-shot learning
   • Prompt injection risks
   • Output formatting and guardrails
   • Prompt Flow pipelines
4. Build RAG-Based Generative Applications
   • Embeddings and vector search
   • Azure AI Search integration
   • Chunking strategies
   • Retrieval optimization
   • RAG vs fine-tuning trade-offs
5. Monitoring, Safety & Optimization
   • Azure AI Content Safety
   • Hallucination mitigation
   • Cost optimization strategies
   • Latency optimization
   • Logging and monitoring

Labs:

• Deploy and test Azure OpenAI model
• Build prompt-driven application
• Create embedding index with Azure AI Search
• Implement RAG chatbot
• Configure safety and monitoring

• What is Agentic AI
• Tool usage & function calling
• Planning and reasoning loops
• Memory design patterns
• Single vs multi-agent systems

1. Tool Calling & Function Integration
   • Azure OpenAI function calling
   • API integration patterns
   • Structured tool invocation
   • Error handling & retries
   • Observability design
2. Agent Orchestration Frameworks
   • Semantic Kernel architecture
   • LangChain with Azure
   • Plugin architecture
   • State and session management

3. Multi-Agent Collaboration Patterns
   • Supervisor agent pattern
   • Task decomposition
   • Critic agent pattern
   • Debate and refinement loops
4. Enterprise Agent Architecture
   • Secure deployment architecture
   • Scaling agent workloads
   • Cost control strategies
   • Governance & audit logging

Labs:

• Build tool-calling agent
• Integrate external API tool
• Implement memory store
• Develop multi-agent workflow
• Deploy scalable agent API

1. Azure AI Language Services Overview
   • Text analytics capabilities
   • Sentiment analysis
   • Key phrase extraction
   • Named Entity Recognition (NER)
2. Custom Text Classification
   • Custom classification projects
   • Training datasets
   • Evaluation metrics
   • Deployment endpoints
3. Conversational Language Understanding
   • Intent recognition
   • Entity extraction
   • Conversational flows
   • Bot integration
4. Speech & Multimodal NLP
   • Speech-to-text
   • Text-to-speech
   • Translation services
   • Integration with GenAI workflows
5. Enterprise NLP Architecture
   • Scaling NLP workloads
   • Monitoring & observability
   • Cost management
   • Responsible AI compliance

Labs:

• Perform sentiment & entity analysis
• Train custom classifier
• Build conversational language model
• Integrate speech services
• Deploy NLP solution

1. Azure AI Vision Fundamentals
   • Image analysis
   • Object detection
   • Tagging & classification
   • OCR capabilities
2. Custom Vision Model Development
   • Dataset preparation
   • Training custom models
   • Model evaluation
   • Deployment endpoints
3. Face & Video Analysis
   • Face detection concepts
   • Video indexing
   • Responsible AI considerations
4. Multimodal AI Integration
   • Vision + LLM integration
   • Image captioning
   • Visual question answering
5. Enterprise Vision Architecture
   • Scaling vision workloads
   • Latency optimization
   • Edge deployment considerations
   • Security & compliance

Labs:

• Analyze images using Vision API
• Build object detection model
• Implement OCR pipeline
• Train custom vision classifier
• Integrate Vision with GenAI
• Deploy scalable vision solution

1. Azure AI Document Intelligence Overview
   • Form Recognizer concepts
   • Prebuilt models
   • Custom extraction models
   • Layout analysis
2. Custom Document Processing Pipelines
   • Dataset preparation
   • Training custom extraction models
   • Confidence scoring
   • Model evaluation
3. Intelligent Document Automation
   • Invoice processing
   • Contract analysis
   • Compliance workflows
   • Integration with RAG systems
4. Enterprise Governance & Production Deployment
   • Secure data pipelines
   • PII handling
   • Key Vault integration
   • Monitoring & cost management

Capstone: End-to-End Intelligent AI Solution

• Design enterprise architecture
• Implement RAG + Agent + Document Intelligence
• Deploy to Azure
• Monitoring and evaluation
• Architecture presentation

Labs:

• Extract structured data from forms
• Train custom document model
• Build automated invoice processor
• Integrate extraction with GenAI
• Deploy end-to-end solution

After completing this program, participants will be able to:

• Build enterprise-grade Generative AI applications
• Design and deploy Agentic AI systems
• Implement NLP and Vision solutions at scale
• Develop intelligent document automation systems
• Apply governance, security, and Responsible AI frameworks
• Optimize cost and performance for Azure AI workloads
• Successfully prepare for AI-102 certification