AI-200: cloud development for AI solutions

Objectives and progression

Six modules with fictional APS and project cases, explained practice, and an internal assessment of 24 decisions in 60 minutes. Covers the four domains through data, delivery, and operational contracts. AI-200 syllabus inspected on 2026-09-30; the guide states an update on 2026-05-05 without explicitly stating an objectives effective date. Official exam of 120 minutes and score 700/1000, not equivalent to 70%, with no fixed question count inferred. Official Portuguese is Brazil.

Audience: Cloud developers, APS teams, and technical application and data managers.

Prerequisites: Python, SQL, containers, APIs, and Azure fundamentals.

300 estimated study minutes

• Deploy reproducible versions and diagnose startup, access, and routing.
• Design access and synchronization around partitions, guarantees, and operation cost.
• Balance retrieval, connections, and freshness without losing isolation.
• Build asynchronous processing that tolerates failures and redeliveries.
• Coordinate configuration changes and retain controlled access and correlation.
• Interpret metrics and traces with explicit denominators, windows, and coverage.

Modules

1. Containers, artifacts, and revisions
2. Cosmos DB, vectors, and change feed
3. PostgreSQL, vectors, and caching
4. Messages, events, and Functions
5. Secrets, configuration, and context
6. Telemetry, KQL, and diagnosis

Continue learning

Microsoft Azure

• AZ-204 — Azure Developer
• Azure AI Apps and Agents Developer Associate
• Microsoft Azure Database Administrator

References and version

AI-200 current guide updated2026-05-05; Azure technical documentation2026-09-30

• AI-200 current objectives; guide updated2026-05-05 · 2026-09-30
• Azure AI Cloud Developer Associate format · 2026-09-30
• Image tags digests and reproducible deployments · 2026-09-30
• ACR Tasks source base-image and timer triggers · 2026-09-30
• App Service container settings and secrets · 2026-09-30
• Container Apps revisions and traffic · 2026-09-30
• Container Apps KEDA scaling · 2026-09-30
• Startup readiness and liveness probes · 2026-09-30
• AKS kubelet image access and ABAC registry exception · 2026-09-30
• Container and replica storage lifetime · 2026-09-30
• Partition key and point-read identity · 2026-09-30
• Cosmos vector policies dimensions and distance · 2026-09-30
• Change feed processing and at-least-once delivery · 2026-09-30
• Latest-version and all-versions-and-deletes modes and SDK support · 2026-09-30
• Cosmos indexing policies and RU tradeoffs · 2026-09-30
• Consistency guarantees and tradeoffs · 2026-09-30
• Managed PostgreSQL vector indexes · 2026-09-30
• PgBouncer transaction pooling and compatibility · 2026-09-30
• Azure Managed Redis data and vector operations · 2026-09-30
• Semantic caching workflow · 2026-09-30
• Cache expiration and invalidation · 2026-09-30
• Service Bus locks and settlement · 2026-09-30
• Dead-letter handling and diagnosis · 2026-09-30
• Queues topics and independent subscriptions · 2026-09-30
• MessageId duplicate-detection window · 2026-09-30
• Event Grid delivery retries and ordering · 2026-09-30
• Event routing filters · 2026-09-30
• Functions triggers and bindings · 2026-09-30
• Python async and blocking I/O · 2026-09-30
• Functions reliable background work · 2026-09-30
• Secret rotation and consumer coordination · 2026-09-30
• Key Vault data and management roles · 2026-09-30
• Immutable configuration snapshots · 2026-09-30
• Python provider refresh and sentinel · 2026-09-30
• Distributed application observability · 2026-09-30
• Telemetry sampling and interpretation · 2026-09-30
• KQL conditional counts applicable to Azure Monitor · 2026-09-30
• KQL percentile estimates applicable to Azure Monitor · 2026-09-30
• Bounded retries and non-idempotent effects · 2026-09-30
• Microsoft scaled exam scores · 2026-09-30
• Service selector endpoints and port diagnosis · 2026-09-30
• pgvector upstream distance operators filtering and iterative scans · 2026-09-30
• Psycopg3 parameter binding · 2026-09-30
• Distributed context propagation · 2026-09-30