AI-200 Tests Whether You Can Build Containerized, Data-Driven AI Apps on Azure
The Microsoft AI-200 exam, the successor to the retired AZ-204, rewards developer judgment: choosing container hosts, using Cosmos DB, PostgreSQL and Redis for vector and AI workloads, wiring services with messaging and Functions, and securing and monitoring the result.
Check Your Readiness →Most candidates understand Microsoft Certified: Azure AI Cloud Developer Associate (AI-200) concepts — and still fail. This exam tests how you apply knowledge under pressure.
AI data services (about 25-30%) is the largest domain, and containers, service integration and security and monitoring each carry about 20-25%. Most items ask you to pick the right host, data service or messaging pattern for a stated requirement, then secure it without secrets in code.
Increase the retry loop in the consumer so it keeps trying the same message until it succeeds.
Let Service Bus move messages that exceed the maximum delivery count (10 by default) to the dead-letter queue, then inspect and reprocess or discard them. Endless retries block the queue and hide poison messages.
Deploy to an AKS cluster and configure the cluster autoscaler.
Use Azure Container Apps with a KEDA scale rule on the queue length, which scales replicas out with load and down to zero when idle, without managing a Kubernetes cluster. Choose AKS only when full Kubernetes control is required.
Scale up the model deployment so every request is answered faster.
Add semantic caching in Azure Managed Redis: store embeddings of previous prompts with their responses and return a cached answer when a new prompt is similar enough. It cuts latency and model cost for repeated questions.
Service Bus is a message broker for reliable, ordered (with sessions) and transactional delivery of commands and work items, with dead-letter queues. Event Grid routes lightweight event notifications to subscribers using push delivery with filters and retries, so it fits reactive scenarios rather than work queues.
Cosmos DB offers five levels: strong, bounded staleness, session, consistent prefix and eventual, with session as the default. Stronger levels cost more RUs and latency, and strong consistency is not available with multi-region writes.
In the default latest-version mode, the Cosmos DB change feed records inserts and updates but not deletes. A common pattern is a soft-delete flag with a TTL so downstream processors see the change before the item expires.
Embedding keys or connection strings in code or app settings creates secrets that leak and need rotation. Assign a managed identity, grant it RBAC roles on the target service and use Key Vault references for secrets that cannot be avoided.
IVFFlat builds its clusters from existing rows, so it should be created after the table holds representative data. HNSW gives better speed-recall trade-offs and can be built at any time, but it uses more memory and builds more slowly.
Strengthen weak areas with exam-style practice questions and detailed explanations.
Simulate the real exam experience and assess your readiness under timed conditions.
Review key concepts, objectives, and exam topics in one place.
Get personalized explanations, learning recommendations, and instant answers.
Follow a structured learning path designed to help you prepare efficiently.
Find out which AI-200 skills area, from AI data services to containers, is costing you the most points.