Microsoft Azure AI Cloud Developer (AI-200) Exam Prep

What's inside

1. Level 1: The AI-200 Exam: Format, Domains and Study PlanFree
   [Exam orientation] Covers what AI-200 (Microsoft Certified: Azure AI Cloud Developer Associate, the replacement for AZ-204) tests and how. That means the four domains of the skills measured as of October 7, 2026 and their weightings, including the clash between 30–35% and 25–30% for the data domain. It also covers the 120-minute time limit set in the course brief (confirm it on Microsoft's exam page, because some third-party sites list a different length), the 700 pass mark, possible interactive items, free renewal, and the fact that Microsoft's study guide is free on Microsoft Learn and is not reproduced here. This course is independent of Microsoft, and it cannot promise you a pass.
2. Level 2: Azure Container Registry: Build, Store and Push ImagesFree
   [Develop containerized solutions on Azure – Implement container application hosting] Teaches how to create an Azure Container Registry, choose a service tier, authenticate with az acr login or Microsoft Entra identities, and push a Dockerized Python app so it is stored in the registry.
3. Level 3: Versioning and Managing Images in ACR
   [Develop containerized solutions on Azure – Implement container application hosting] Covers how to version and manage container images in Azure Container Registry: tags compared with digests, why a reused 'latest' tag is risky, locking tags, deleting and purging old images, and listing repositories and manifests.
4. Level 4: ACR Tasks: Quick Builds and Runs in the Cloud
   [Develop containerized solutions on Azure – Implement container application hosting] Teaches how to build and run images with Azure Container Registry Tasks, starting with quick tasks: az acr build to build in Azure without a local Docker engine, and az acr run to run commands against the registry.
5. Level 5: Automated and Multi-Step ACR Tasks
   [Develop containerized solutions on Azure – Implement container application hosting] Covers automated Azure Container Registry Tasks: tasks triggered by a source commit, by a base image update or on a schedule, plus multi-step tasks defined in a YAML file that build, test and push images.
6. Level 6: Deploying Containers to Azure App Service
   [Develop containerized solutions on Azure – Implement container application hosting] Teaches how to deploy a custom container from Azure Container Registry to Azure App Service on Linux: pulling the image with a managed identity, setting the container port, and enabling continuous deployment when the image changes.
7. Level 7: App Service Environment Variables and Secrets
   [Develop containerized solutions on Azure – Implement container application hosting] Covers how to configure Azure App Service to supply environment variables and secrets to a container: app settings, connection strings, deployment slot settings, and Key Vault references resolved through a managed identity.
8. Level 8: Azure Container Apps: Environments and Deployment
   [Develop containerized solutions on Azure – Implement container-orchestrated solutions] Teaches how to deploy applications to Azure Container Apps, including environment configuration: the Container Apps environment as a shared boundary for networking and logging, ingress, environment variables, secrets and pulling images from ACR.
9. Level 9: Container Apps Revisions and Traffic Splitting
   [Develop containerized solutions on Azure – Implement container-orchestrated solutions] Covers revision management in Azure Container Apps: what creates a new revision and what does not, single compared with multiple revision mode, traffic weights, revision labels, and rolling back.
10. Level 10: KEDA Event-Driven Scaling in Container Apps
    [Develop containerized solutions on Azure – Implement container-orchestrated solutions] Teaches how to implement event-driven scaling with Kubernetes Event-driven Autoscaling (KEDA) in Container Apps: HTTP, TCP and custom scale rules, such as a Service Bus queue scaler, minimum and maximum replicas, scale to zero, and authenticating scalers.
11. Level 11: Deploying to AKS with Manifest Files
    [Develop containerized solutions on Azure – Implement container-orchestrated solutions] Covers how to deploy and manage applications on Azure Kubernetes Service (AKS) with manifest files: attaching ACR to the cluster, and writing Deployment, Service, ConfigMap and Secret YAML that you apply, update and roll back with kubectl.
12. Level 12: Troubleshooting AKS and Container Apps
    [Develop containerized solutions on Azure – Implement container-orchestrated solutions] Teaches how to monitor and troubleshoot solutions on AKS and Container Apps by inspecting logs, events and end-to-end connectivity: kubectl logs, describe and get events, az containerapp logs show, Log Analytics tables, and tracing a request from ingress to pod.
13. Level 13: Cosmos DB for NoSQL: Connecting with the Python SDK
    [Develop AI solutions by using Azure data management services – Azure Cosmos DB for NoSQL] Covers how to connect to Azure Cosmos DB for NoSQL with the azure-cosmos Python SDK: CosmosClient with Microsoft Entra ID or keys, the account, database, container and item hierarchy, choosing a partition key, and point reads and writes.
14. Level 14: Querying Cosmos DB and Measuring Request Units
    [Develop AI solutions by using Azure data management services – Azure Cosmos DB for NoSQL] Teaches how to run queries with the SDK and read their cost in Request Units (RUs): the NoSQL query syntax, parameterized queries, single-partition compared with cross-partition queries, paging, and the request-charge header.
15. Level 15: Cosmos DB Indexing Policies for Performance
    [Develop AI solutions by using Azure data management services – Azure Cosmos DB for NoSQL] Covers how to optimize query performance and RU consumption with indexing policies: included and excluded paths, consistent compared with none indexing mode, range, spatial and composite indexes, and the trade-off between write cost and query cost.
16. Level 16: Consistency Levels and Their RU Trade-offs
    [Develop AI solutions by using Azure data management services – Azure Cosmos DB for NoSQL] Teaches the five consistency levels (strong, bounded staleness, session, consistent prefix, eventual): how each affects read RUs, latency and availability, why session is the default, and how a client can relax consistency per request.
17. Level 17: Storing Embeddings in Cosmos DB for NoSQL
    [Develop AI solutions by using Azure data management services – Azure Cosmos DB for NoSQL] Covers how to store and retrieve embeddings: enabling vector search on the account, defining a container vector embedding policy (path, data type, dimensions, distance function), choosing a flat, quantizedFlat or diskANN vector index, and excluding vector paths from the regular index.
18. Level 18: Vector Similarity Search in Cosmos DB
    [Develop AI solutions by using Azure data management services – Azure Cosmos DB for NoSQL] Teaches how to run a vector similarity search for semantic retrieval with the VectorDistance system function: ORDER BY with TOP N, combining a vector search with WHERE filters, reading similarity scores, and feeding the results into a retrieval-augmented generation (RAG) prompt.
19. Level 19: The Change Feed Processor
    [Develop AI solutions by using Azure data management services – Azure Cosmos DB for NoSQL] Covers how to implement a change feed processor to detect and handle new or updated items: the monitored container, the lease container, the handler delegate, instances and load balancing, what latest version mode does and does not capture, and Python options such as the change feed pull model and the Azure Functions Cosmos DB trigger.
20. Level 20: Connecting to Azure Database for PostgreSQL
    [Develop AI solutions by using Azure data management services – Azure Database for PostgreSQL] Teaches how to connect to and query Azure Database for PostgreSQL flexible server through SDKs: psycopg with Microsoft Entra token authentication or password authentication, required TLS, firewall and private access rules, and parameterized queries from Python.
21. Level 21: PostgreSQL Schema Design and Data Types for AI
    [Develop AI solutions by using Azure data management services – Azure Database for PostgreSQL] Learn to model schemas for AI workloads on Azure Database for PostgreSQL: design tables that keep source text, metadata and embeddings together, and choose suitable data types such as vector(n), jsonb, text, timestamptz and identity keys, with the reasoning the exam tests behind each choice.
22. Level 22: pgvector Indexing: Latency vs Compute Overhead
    [Develop AI solutions by using Azure data management services – Azure Database for PostgreSQL] Learn indexing strategies that cut query latency and reduce pgvector compute overhead: exact scans compared with HNSW and IVFFlat approximate indexes, operator classes that must match the distance operator, B-tree and GIN indexes on filter columns, and where the DiskANN extension fits (it is in public preview at the time of writing).
23. Level 23: Sizing PostgreSQL for Vector Workloads
    [Develop AI solutions by using Azure data management services – Azure Database for PostgreSQL] Learn to configure compute, memory and storage resources in Azure Database for PostgreSQL for vector workloads: compute tiers and vCores, why HNSW indexes need to fit in memory, maintenance_work_mem for index builds, and storage size and IOPS.
24. Level 24: PostgreSQL Vector Search and RAG with Metadata Filters
    [Develop AI solutions by using Azure data management services – Azure Database for PostgreSQL] Learn to run vector similarity search end to end: store embeddings from an Azure OpenAI model, do semantic retrieval with ORDER BY distance and LIMIT, and build retrieval-augmented generation (RAG) patterns that use metadata filters to scope the results.
25. Level 25: PostgreSQL Connection Optimization
    [Develop AI solutions by using Azure data management services – Azure Database for PostgreSQL] Learn connection optimization to improve throughput and minimize latency: client-side pooling with psycopg_pool, the built-in PgBouncer on Azure Database for PostgreSQL, Microsoft Entra token authentication, and keeping connections close to the app.
26. Level 26: Azure Managed Redis: Caching, Expiration, Invalidation
    [Develop AI solutions by using Azure data management services – Integrate Azure Managed Redis in AI solutions] Learn Azure Managed Redis data operations in Python with redis-py: the cache-aside pattern, setting expiration with TTLs, eviction policies, and invalidating stale entries when the source data changes.
27. Level 27: Azure Managed Redis Vector Indexing
    [Develop AI solutions by using Azure data management services – Integrate Azure Managed Redis in AI solutions] Learn to use vector indexing in Azure Managed Redis to enable similarity search: enable the RediSearch module at creation time (it requires the Enterprise clustering policy and the NoEviction policy), create a vector index, and run KNN queries for patterns such as semantic caching.
28. Level 28: Service Bus Queues and Messages
    [Connect to and consume Azure services – Develop event- and message-based AI solutions] Learn to queue and process back-end operations with Azure Service Bus queues using the azure-servicebus Python SDK: sending and receiving messages, peek-lock compared with receive-and-delete, message settlement, sessions and duplicate detection.
29. Level 29: Service Bus Topics and Subscriptions
    [Connect to and consume Azure services – Develop event- and message-based AI solutions] Learn publish-subscribe with Azure Service Bus topics and subscriptions: fan-out to several consumers, SQL and correlation filters, rule actions, and setting application properties for routing.
30. Level 30: Service Bus Dead-Letter Queue Handling
    [Connect to and consume Azure services – Develop event- and message-based AI solutions] Learn dead-letter queue handling in Azure Service Bus: how messages reach the dead-letter queue (max delivery count, TTL expiry, filter evaluation errors, explicit dead-lettering), and how to inspect, fix and resubmit them.
31. Level 31: Event Grid Custom Events
    [Connect to and consume Azure services – Develop event- and message-based AI solutions] Learn to build event-driven workflows with Azure Event Grid custom events: custom topics, the CloudEvents and Event Grid schemas, publishing with the azure-eventgrid SDK, and subscribing webhooks, Functions and queues as handlers.
32. Level 32: Event Grid Filters
    [Connect to and consume Azure services – Develop event- and message-based AI solutions] Learn to route only the events each handler needs with Azure Event Grid subscription filters: event type filters, subject begins-with and ends-with filters, and advanced filters on data fields.
33. Level 33: Event Grid Retries and Dead-Lettering
    [Connect to and consume Azure services – Develop event- and message-based AI solutions] Learn how Azure Event Grid delivers events with retries: the retry policy (maximum delivery attempts and event time-to-live), which response codes are retried, dead-lettering undelivered events to Blob Storage, and making handlers idempotent.
34. Level 34: Azure Functions Triggers: Serverless APIs
    [Connect to and consume Azure services – Develop and implement Azure Functions] Learn to build serverless APIs with the Azure Functions Python v2 programming model: HTTP triggers declared with decorators, routes and auth levels, plus timer, Service Bus and Event Grid triggers that call your AI back end.
35. Level 35: Azure Functions Input and Output Bindings
    [Connect to and consume Azure services – Develop and implement Azure Functions] Learn how triggers and bindings work together: input and output bindings for Azure Cosmos DB, Service Bus and Blob Storage, connection settings that point to app settings or identity-based connections, and when a binding is better than calling the SDK directly.
36. Level 36: Configuring Function Apps
    [Connect to and consume Azure services – Develop and implement Azure Functions] Learn to configure function apps: hosting plans (Flex Consumption, Premium, Dedicated and container hosting), required app settings such as AzureWebJobsStorage and APPLICATIONINSIGHTS_CONNECTION_STRING, host.json, managed identity, and scale and timeout limits.
37. Level 37: Deploying Function Apps
    [Connect to and consume Azure services – Develop and implement Azure Functions] Learn to deploy function apps: func azure functionapp publish, OneDeploy and zip deployment, remote build for Python, deploying from CI/CD, and verifying that a deployment worked.
38. Level 38: Key Vault: Retrieving Secrets Securely
    [Secure, monitor, and troubleshoot Azure solutions – Implement secure Azure solutions] Learn to retrieve secrets from Azure Key Vault with the azure-keyvault-secrets SDK and DefaultAzureCredential, choose between Azure RBAC and access policies, and reference Key Vault from App Service, Container Apps and Functions settings.
39. Level 39: Key Vault Secret Rotation
    [Secure, monitor, and troubleshoot Azure solutions – Implement secure Azure solutions] Learn secret rotation with Azure Key Vault: secret versions, expiry dates, near-expiry events published to Event Grid that trigger a rotation function, and writing apps that pick up new versions without redeploying.
40. Level 40: Azure App Configuration
    [Secure, monitor, and troubleshoot Azure solutions – Implement secure Azure solutions] Learn to store and retrieve app configuration information with Azure App Configuration: keys, labels for environments, feature flags, Key Vault references, and dynamic refresh with a sentinel key from the Python provider.
41. Level 41: App Configuration: Key Vault refs, labels and refresh
    [Secure, monitor, and troubleshoot Azure solutions] Learners take the basic Azure App Configuration store one step further. They use the azure-appconfiguration-provider load() call with key filters and labels per environment, resolve Key Vault references with a managed identity, read feature flags, and set up dynamic refresh with a sentinel key, then compare the result with putting secrets directly in App Configuration.
42. Level 42: OpenTelemetry Basics: Traces, Spans and Azure Monitor
    [Secure, monitor, and troubleshoot Azure solutions] Learners trace a distributed system with the OpenTelemetry SDKs. They learn what traces, spans, span kinds and context are, then send telemetry from a Python app to Application Insights with the Azure Monitor OpenTelemetry Distro and a connection string.
43. Level 43: OpenTelemetry: Custom Spans, Attributes and Sampling
    [Secure, monitor, and troubleshoot Azure solutions] Learners add manual instrumentation to their OpenTelemetry tracing. They wrap AI pipeline steps such as embedding calls, vector queries and LLM calls in custom spans, record attributes, events, exceptions and span status, and weigh the cost and visibility trade-offs of sampling.
44. Level 44: End-to-End Tracing Across Services and Messaging
    [Secure, monitor, and troubleshoot Azure solutions] Learners trace a single request as it crosses HTTP calls, Azure Service Bus messages, Azure Functions and container apps. The level covers W3C Trace Context (traceparent) propagation, the operation ID, and reading the result in the Application Insights application map and end-to-end transaction view.
45. Level 45: KQL Fundamentals for Application Logs
    [Secure, monitor, and troubleshoot Azure solutions] Learners write KQL queries against the logs in a Log Analytics workspace, mainly the AppRequests, AppDependencies, AppTraces and AppExceptions tables. They use the tabular operators where, project, extend, summarize, count, top, order by and ago() to answer everyday troubleshooting questions.
46. Level 46: KQL for Metrics, Performance and Time Series
    [Secure, monitor, and troubleshoot Azure solutions] Learners use KQL to analyze metrics and performance. The level covers bin() time buckets, percentiles() for latency, make-series and render timechart, and joining requests to dependencies through the operation ID to find out which downstream call is slowing an AI endpoint.
47. Level 47: Troubleshooting Drill: Containers, Traces and KQL
    [Secure, monitor, and troubleshoot Azure solutions] This level ties the domain together in exam-style troubleshooting scenarios. Learners query container logs (ContainerAppConsoleLogs_CL, ContainerAppSystemLogs_CL, ContainerLogV2, KubeEvents) together with Application Insights traces, and connect what they find to Key Vault and App Configuration misconfigurations. They also practise strategies for scenario questions that ask for the best answer.
48. Level 48: Final Mock Exam (50 Questions, 120 Minutes) Timed mock
    [All domains] This timed mock exam has 50 original scenario and code-reading questions in 120 minutes, drawn from all four AI-200 domains in proportion to their weightings. It tests whether the learner is ready for the real exam, using a 70% pass mark and explaining why each distractor is wrong. These are not real exam questions.

Access

The first 2 levels are free with a free account. Every level, the podcast edition and the AI tutor come with All Access at £4.99/month or any Creator plan — see pricing.

Start free →