1. What is Azure AI?
1.1 Agenda
Estimated reading time: ~8 minutes
Learning Outcomes
- Define Azure AI and its position within the broader Azure cloud platform
- Explain the five key benefits of using managed (được quản lý bởi Microsoft) AI services
- Describe the role of Azure AI Foundry as the unified (thống nhất) development platform
- Articulate the difference between prebuilt AI services and custom ML
1.2 Glossary
| Term | Quick Explanation |
|---|---|
| Azure | Nền tảng điện toán đám mây (cloud computing) của Microsoft — cung cấp hạ tầng, dữ liệu và ứng dụng như dịch vụ. |
| Managed Service | Dịch vụ mà Microsoft vận hành và bảo trì hạ tầng — người dùng chỉ cần gọi API, không cần tự dựng server. |
| API | Application Programming Interface — giao diện lập trình cho phép ứng dụng của bạn gọi tính năng AI qua HTTP request. |
| Azure AI Foundry | Nền tảng tập trung để xây dựng, triển khai (deploy) và quản lý ứng dụng AI trên Azure (trước đây là Azure AI Studio). |
| Prebuilt Model | Mô hình AI đã được Microsoft huấn luyện (train) sẵn — người dùng gọi trực tiếp qua API mà không cần training lại. |
| Responsible AI | Bộ nguyên tắc và công cụ đảm bảo AI được phát triển và triển khai một cách công bằng, an toàn và có trách nhiệm. |
2. Problem Statement
Building AI from scratch (từ đầu) requires:
- Deep ML expertise — Data scientists who can design, train, and tune models.
- Massive infrastructure — GPUs, storage, orchestration (điều phối) systems.
- Continuous maintenance — Models degrade (suy giảm hiệu năng) over time and need retraining.
- Security & compliance overhead (gánh nặng tuân thủ) — Enterprise-grade (cấp doanh nghiệp) data protection and regulatory requirements.
Azure AI removes all four barriers through managed services — Microsoft handles the infrastructure, model maintenance, security, and compliance. Your team integrates AI via API calls.
3. What is Azure AI?
3.1 Definition
Azure AI is Microsoft's collection of artificial intelligence services, tools, and infrastructure that enables organizations to add AI capabilities to their applications — without requiring deep data science expertise or building AI models from scratch.
3.2 Position in the Azure Ecosystem
3.3 Azure AI Foundry
Azure AI Foundry (formerly (trước đây là) Azure AI Studio) is the central workbench (bàn làm việc trung tâm) for building AI applications on Azure. It provides:
- Model Catalog (Danh mục mô hình): Access to models from Microsoft, OpenAI, Meta, Hugging Face, and others.
- Prompt Flow: Visual orchestration (điều phối trực quan) of AI workflows.
- Evaluation & Safety tools: Test model quality and apply content filters (bộ lọc nội dung) before deployment.
- Agent capabilities: Build AI agents that act autonomously (hành động tự chủ) on multi-step tasks.
4. Why Use Azure AI?
4.1 Five Key Benefits
| Benefit | What It Means |
|---|---|
| Scalability (Khả năng mở rộng) | Scale from a prototype (nguyên mẫu) to millions of requests without re-architecting (thiết kế lại) your system. |
| Security & Compliance | Built-in encryption (mã hóa), role-based access control (RBAC), and compliance with GDPR, HIPAA, ISO 27001. |
| Integration | Native connectivity (kết nối gốc) to Azure Storage, Azure SQL, Azure Functions, and 300+ other Azure services. |
| Speed to Market (Tốc độ ra thị trường) | Add AI features in days via API — no need to train models from scratch. |
| Responsible AI | Microsoft's built-in fairness (công bằng), transparency (minh bạch), and accountability (trách nhiệm) tools across all services. |
4.2 Managed vs. Custom: When to Choose Each
| Factor | Prebuilt AI Services | Custom ML (Azure ML) |
|---|---|---|
| Time to implement | Hours to days | Weeks to months |
| ML expertise needed | Minimal | Significant |
| Customization | Limited | Full control |
| Data required | None (use pretrained (đã huấn luyện trước) models) | Large labeled dataset needed |
| Best for | Common tasks (sentiment, OCR, translation) | Domain-specific (chuyên biệt theo lĩnh vực) problems |
5. Azure AI Service Categories
Workshop 2 covers the core prebuilt AI services, organized by the AI workload they serve:
| Category | Azure Service | Workload |
|---|---|---|
| Language | Azure AI Language | NLP — text analytics, Q&A, classification |
| Language | Azure AI Translator | NLP — real-time translation |
| Speech | Azure AI Speech | Speech-to-text, TTS, speaker recognition |
| Vision | Azure AI Vision | Image analysis, OCR, spatial analysis |
| Vision | Azure AI Document Intelligence | Document-specific OCR + field extraction |
| Safety | Azure AI Content Safety | Detect & filter harmful content |
| Multimodal | Azure AI Content Understanding | Extract insights from multimodal (đa phương thức — text, ảnh, audio, video) unstructured data |
Each service is covered in detail in Workshop 2.2 — Azure AI Services.
6. Discussion Questions
Q1 — Build vs. Buy: A fintech (công nghệ tài chính) startup wants to add sentiment analysis to customer feedback. Their CTO proposes training a custom ML model; their engineer proposes using Azure AI Language's prebuilt sentiment API. What factors should drive this decision? Under what conditions would the custom approach be justified (được biện minh)?
Q2 — The Managed Service Trade-off: Using a managed service means Microsoft controls the underlying (bên dưới) model. If Microsoft updates the model, your application's behavior may change without your intervention (can thiệp). What governance (quản trị) practices should your team implement to detect and respond to such changes?
Q3 — Responsible AI as a Feature: A company pitches (quảng bá) to an enterprise client: "We use Azure AI, which has built-in Responsible AI." Is this claim sufficient (đủ) to satisfy (đáp ứng) the client's AI governance requirements? What additional (bổ sung) responsibilities remain with the application developer, regardless of (bất kể) the platform?
Made by Anh Tu - Share to be share