Section 1
Planning & Management
Preview
Preview: Plan and manage an Azure AI solution
- 4 of 16 lesson topics
- 1 overview segment
- 3 core concepts
- 2 exam tips
Lesson Topics
- Choosing Azure AI Search or Azure Cosmos DB
- Selecting Models for Low-Latency Chat
- Reasoning Model Selection
- Vector and Hybrid Retrieval Planning
Overview
This section covers planning choices that affect Azure AI solution cost, latency, scalability, and reliability.
Core Concepts
- **Azure AI Search vs. Azure Cosmos DB:** Azure AI Search is optimized for search, retrieval, ranking, vector search, hybrid search, and filtering. Azure Cosmos DB is better when the application also needs operational data storage, transactional consistency, or document database patterns.
- **Model Selection:** GPT-4o-class models are commonly chosen for low-latency conversational and multimodal experiences. Reasoning models are better fits for complex logic, coding, and multi-step analysis when the extra latency and cost are justified.
- **Reasoning Tokens and Limits:** Reasoning models can consume internal reasoning tokens that count toward output billing and token limits. Use the supported token-limit parameter for the model/API rather than assuming older chat-completion parameters apply everywhere.
Exam Tips
- Read latency, scale, filtering, and data consistency requirements before choosing Azure AI Search or Azure Cosmos DB.
- Do not assume a reasoning model is always better; it is justified when the task requires deeper analysis and can tolerate extra latency/cost.