Can you ship an AI-backed backend that is deployable, queryable, event-driven, observable, and secure — not just working on your laptop?

Prepare for Microsoft AI-200 (Developing AI Cloud Solutions on Azure) with 600 original questions across six full-length mock exams of 100 items each. Every exam is blueprint-weighted, so each one is a coherent rehearsal of the whole blueprint rather than a random topic dump.

This is a developer exam about the back end of AI systems. It expects working judgment about container images and registries, App Service, Container Apps and KEDA scaling, Azure Kubernetes Service manifests, Cosmos DB vectors and change feed, PostgreSQL with pgvector, Azure Managed Redis caching and vector search, Service Bus and Event Grid messaging, Azure Functions triggers and bindings, Key Vault, App Configuration, OpenTelemetry and KQL.

WHAT YOU GET

• Six original 100-question mock exams

• Blueprint weighting on every exam: 24 / 30 / 24 / 22 across the four published domains

• Mixed item types: 76 single-choice, 16 multi-select, 8 ordering items per exam

• Four three-question case studies inside every exam

• An explanation for every option plus a case-specific overall rationale

• Local study-map skill anchors and the official study-guide reference on every item

• Offline Markdown review copies alongside instructor CSV files

• A reproducible ZIP, a validation report, and automated structural tests

HOW THE ITEM MIX MATCHES THE REAL EXAM

The live role-based exam is not a wall of identical single-answer items. Microsoft's exam experience guidance describes 40–60 questions in a 120-minute appointment with possible interactive components, and candidates routinely see multi-select, drag-and-drop or ordering, hot-area, case-study, and adaptive problem-solution formats.

These mock exams reproduce what a CSV player can score honestly:

• Single-choice items with one best answer.

• Multi-select items that state the required count ("Select two") and accept a comma-separated key.

• Ordering items rendered as four candidate sequences, so sequencing judgment is tested without a drag-and-drop UI.

• Case studies: a shared scenario followed by three scored questions, exactly as a case-study block behaves in the exam.

What is deliberately not reproduced, and is stated in every export: hot-area image selection, drag-and-drop UI, hands-on labs and interactive components, adaptive problem-solution (yes/no) sets, and the exact live item count.

A NOTE ON THE PUBLISHED WEIGHTS

The study guide's domain heading says the data-management domain is 25–30%, while its own skills-at-a-glance list says 30–35%, and the change log records that this domain's share increased. This course allocates exactly 30% to that domain — the only value inside both statements — and keeps the other three inside 20–25%. Verify the current guide before you schedule.

THE SIX EXAMS

1. RETAIL RECOMMENDATION PLATFORM

Registry tagging and ACR Tasks, App Service settings and Key Vault references, Container Apps ingress and revisions, KEDA queue scaling, AKS manifests and probes, container connectivity; Cosmos DB partition reads, indexing policies, consistency, vector search and change feed; pgvector indexing and pooling; Redis caching and vector search; Service Bus sessions, topics, filters and dead-lettering; Event Grid filters and retries; Functions triggers, bindings and hosting plans; Key Vault, App Configuration, OpenTelemetry and KQL.

2. HEALTHCARE DOCUMENT INTAKE

Immutable artifacts, slot swaps, warm minimum replicas, private DNS, readiness probes; Cosmos DB DiskANN and leases; IVFFlat, vacuum and work memory; metadata-filtered retrieval before selection; connection pooling under serverless bursts; TTL and invalidation; Service Bus prefetch and sessions; system topics; Durable Functions; premium plans and managed identities; data-plane vault access; configuration refresh; time-bounded KQL.

3. PAYMENTS AND FRAUD SCORING

Registry geo-replication, KEDA scaler identity, workload identity federation; Cosmos DB vector policy dimensions and distance functions, transactional batches; HNSW versus IVFFlat choice, EXPLAIN ANALYZE, prepared statements; duplicate detection and scheduled messages; identity-based event delivery; tail sampling, platform metric alerts, materialized aggregations; private endpoints and least privilege.

4. FACTORY QUALITY INSPECTION

Retention policies, deployment slots, scale-to-zero review workers, manifest reproducibility, in-pod connectivity checks; Cosmos RU charge per write, session consistency, change feed leases; pgvector schema design, compute sizing, connection pooling; Service Bus dead-lettering; Event Grid subject filters and retries; durable orchestration for rework; threshold configuration and refresh; trace joins and dashboards.

5. CITIZEN DOCUMENT SERVICES

App Service and slot patterns, revision canaries, cool-down tuning, private name resolution; Cosmos change feed, indexing policies, eventual consistency; post-filtering pitfalls, external poolers, session state in Redis; peek-lock recovery, idempotency, dead-lettering; system topics; consumption plans and slot deployment; purge protection, labels, Key Vault references; collector routing and request-to-dependency joins.

6. GLOBAL RETAIL CAPSTONE

Multi-region registries, per-workload identity, diagnostic routing, canary release, warm capacity; DiskANN at scale, per-request consistency relaxation, metadata-scoped retrieval, pooler sizing; topic fan-out with filters, idempotent consumers, custom events, durable workflows; tail sampling, materialized aggregation, private endpoints and least privilege.

EACH EXAM CONTAINS

• 12 questions on Develop containerized solutions on Azure

• 15 questions on Develop AI solutions by using Azure data management services

• 12 questions on Connect to and consume Azure services

• 11 questions on Secure, monitor, troubleshoot Azure solutions

COVERAGE THAT IS EXPLICIT ABOUT ITS LIMITS

All four published domains and 30 local study-map skills have mapped practice. Numeric skill IDs are local study aids, not Microsoft objective numbers. Mapping does not mean every sub-objective has equal depth, and the questions have not been psychometrically calibrated against the live exam.

HOW THE CERTIFICATION ACTUALLY WORKS

The exam is proctored and lasts 120 minutes. Microsoft may include interactive components. The passing score is 700 out of 1000, the exam costs $165 USD, it is available in English plus twelve other languages, it is valid for one year with a free renewal assessment, and a retake is possible 24 hours after the first attempt.

HOW PRACTICE DIFFERS FROM THE SCORE

A 100-item exam is more than twice the live item count, so the real 120-minute clock does not map directly. Use 100 minutes for a deliberate pace drill, or 150 minutes for a full review attempt. Any percentage target you set is a study convention. Microsoft reports a scaled score of 700/1000, not a raw percentage, so a practice percentage is not a scaled exam score and no conversion is implied.

Concepts and option definitions recur across exams for reinforcement. Three hundred distinct question stems do not mean three hundred independent concepts or six independent readiness estimates. No score, calibrated difficulty, lab completion, or certification result is guaranteed.

This is a practice-test course, not a lecture series, provisioned lab environment, Azure subscription, exam voucher, or certification award. Instructor CSV files require account-level import preview before publication. All items are original educational content, not actual, recalled, or copied exam questions. This independent course is not affiliated with or endorsed by Microsoft.