Manage, monitor and secure AI systems
AI-103 · Lesson 3 of 15 · 17:09
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Prepare for the AI-103 exam (Azure AI Apps and Agents Developer Associate). In this lesson: quotas and rate limits (tokens per minute, 429 errors, retries and max_tokens), costs and capacity (standard and provisioned throughput, reservations, spillover, Batch, budgets, prompt caching), monitoring (Azure Monitor metrics, diagnostic settings, continuous and scheduled evaluations, model version drift, Azure AI Search indexer health), and security (Entra ID instead of keys, managed identities, agent identities, Foundry roles, private endpoints and virtual networks). Includes three original practice questions.
Preview the topics
- A chat app gets 429 errors during short bursts at peak time, although its average use stays well under the deployment’s limit. What should you do first?
- A team wants to spot a slow drop in the quality of its RAG agent over several months, whatever the traffic looks like. What should they set up?
- A published Foundry agent must read files in a Blob container. You gave Storage Blob Data Reader to the project’s managed identity, but the agent still gets access denied. What’s the fix?