Course Overview
This course is designed to help data scientists, ML engineers, and AI developers establish a robust AI infrastructure using Azure services. Participants will work with Azure Machine Learning, OpenAI, Cognitive Search, DevOps, and containerized environments to develop and operationalize modern AI solutions, including RAG systems and large language model (LLM) deployments.
Duration: 40 hours
Course Objectives
By the end of this course, learners will:
- Understand Azure AI service offerings and architecture components.
- Set up AI-ready infrastructure using Azure ML, Storage, and Compute.
- Deploy and monitor AI models using Azure MLOps pipelines.
- Fine-tune and integrate OpenAI and custom LLMs.
- Build RAG-based applications with Azure Cognitive Search and Blob Storage.
- Apply governance, security, and cost optimization practices in AI projects.
|
Module |
Topics |
Lab Activities |
Duration (hrs) |
| 1. Introduction to AI & Azure Ecosystem | – AI use cases in Azure | – Set up Azure trial |
3 |
| – Azure resource groups, pricing | – Create resource group & AML workspace | ||
| – Key services: AML, OpenAI, Cognitive Search | |||
| 2. Azure Storage for AI | – Blob Storage, ADLS Gen2 | – Upload data to Blob |
3 |
| – Integration with AI pipelines | – Connect data to Azure ML | ||
| – Secure data access | |||
| 3. Compute Infrastructure | – Compute Instances & Clusters | – Create compute cluster |
3 |
| – VM types for training & inferencing | – Train model using compute instance | ||
| – Auto-scaling | |||
| 4. Data Ingestion & Preprocessing | – Pipelines for ingestion | – Build pipeline using ADF |
3 |
| – Azure Data Factory basics | – Label sample dataset | ||
| – Data labeling | |||
| 5. Model Training & Experimentation | – Azure ML SDK/Studio | – Train model using AML |
3 |
| – MLflow integration | – Track runs with MLflow | ||
| – Parameter tuning | |||
| 6. OpenAI & LLM Integration | – Azure OpenAI service | – Call OpenAI API |
4 |
| – Prompt engineering | – Build custom prompt templates | ||
| – Fine-tuning GPT models | |||
| 7. RAG Architectures in Azure | – Cognitive Search with vectors | – Index docs with Cognitive Search |
4 |
| – Vector DBs vs SQL | – Implement RAG using LlamaIndex + OpenAI | ||
| – Building RAG pipelines | |||
| 8. Deployment & Serving | – Real-time vs batch inference | – Deploy model to real-time endpoint |
3 |
| – Endpoints in Azure ML | – Dockerize model for AKS | ||
| – AKS & Container Apps | |||
| 9. MLOps & CI/CD Pipelines | – Azure DevOps basics | – Build YAML pipeline for model lifecycle |
3 |
| – YAML pipelines | – Deploy via DevOps | ||
| – GitHub Actions with Azure ML | |||
| 10. Monitoring, Logging & Drift Detection | – Azure Monitor | – Add logging to endpoint |
3 |
| – Application Insights | – Set up drift detection monitor | ||
| – Model drift & alerts | |||
| 11. Governance & Security | – Role-based access (RBAC) | – Create RBAC policy |
2 |
| – Key Vault | – Secure credentials with Key Vault | ||
| – Compliance & audit | |||
| 12. Cost Management & Optimization | – Cost analysis tools | – Analyze and optimize costs using Azure Calculator |
2 |
| – Reserved instances | |||
| – Workload planning | |||
| 13. Capstone Project | – Build and deploy an AI pipeline using OpenAI + Cognitive Search + Azure ML | – End-to-end lab (from data ingestion to serving via RAG) |
4 |