↓ Skip to main content
Brian Swiger
  1. Certifications/
  2. AI-200 Study Hub/

AI-200 Fact-List Flashcards

Printable fact-list flashcards for AI-200: Developing AI Cloud Solutions on Azure. The front of each card asks a multi-part recall question. The back answers it with the facts the exam turns on: names, defaults, limits, and numbers, plus the traps that the wrong answers rely on.

Created by John Coleman. My friend and colleague John made the original AI-200 Fact-List Flashcards: 34 cards, each one packed with the details that decide exam questions. His cards are the heart of this deck.

Using John's format, the deck has been expanded so that every question in the AI-200 exam simulator is covered by at least one card. Thank you, John, for sharing your work with the community.

Download
#

68Cards
109Simulator questions covered
4Domains, color coded
17Sheets, printed double-sided
Download all 68 flashcards (PDF)

Double-sided, four cards per page · 34 pages · 72 KB

Checked against Microsoft Learn. Every fact on every card was checked against current Microsoft Learn documentation, the same standard as the exam simulator. When a card and the official study guide disagree, the study guide wins.

How to print them
#

  1. Print the PDF double-sided and choose flip on long edge (some printers call it long-edge binding). Each back page is laid out as a mirror image of its front, so every answer lands directly behind its question.
  2. Print at actual size (100%). Actual size keeps the fronts and backs lined up exactly. The pages are US Letter.
  3. Cut along the dashed lines. Each sheet makes four cards.

How to study with them
#

  • Answer every part out loud before you turn the card over. The fronts ask several things at once, the way exam questions combine a service, a setting, and a limit.
  • Sort as you go. Make one pile for the cards you know cold and one for the rest, then shuffle the second pile and go again tomorrow.
  • Study by domain. The colored band across the top of each card shows its exam domain, so you can pull one domain at a time. Domain 2, data management services, carries the most weight on the exam.
  • Pair the cards with the simulator. After a practice run in the exam simulator, the results screen shows your score in each domain. Pull the cards for your lowest domain and study those first.

What the deck covers
#

The cards follow the four official exam domains. Every question in the exam simulator is covered by at least one card.

Domain 1

Develop containerized solutions on Azure

17 cards · 26 of 26 questions covered

Domain 2

Develop AI solutions by using Azure data management services

20 cards · 30 of 30 questions covered

Domain 3

Connect to and consume Azure services

14 cards · 26 of 26 questions covered

Domain 4

Secure, monitor, and troubleshoot Azure solutions

17 cards · 27 of 27 questions covered

Every card
#

Cards are numbered as they are printed. Cards from John’s original deck are marked.

Domain 1 Develop containerized solutions on Azure 17 cards

  1. Card 1: ACR Tasks triggers Original deck

    Name the three ACR Tasks trigger types. Which events are enabled by default? What is NOT a trigger?

  2. Card 2: ACR pull and push roles Original deck

    Which RBAC roles grant image pull and push? What changes on an ABAC-enabled registry?

  3. Card 3: ACR Tasks build types

    Name the three ACR Tasks scenarios. What makes a task rebuild on every commit? What is NOT an ACR Task?

  4. Card 4: ACR tags and digests

    Stable tag, unique tag or digest - which pins one exact image? How do you get an image into ACR securely?

  5. Card 5: App Service pull from ACR

    Two ways App Service can authenticate to ACR. Which one stores no credentials, and what role does it need?

  6. Card 6: App Service app settings

    How does App Service pass configuration to a container? App setting, connection string or Dockerfile - which fits?

  7. Card 7: App Service Key Vault refs

    What does a Key Vault reference look like? What must the app have before it resolves? How fast is a rotated secret picked up?

  8. Card 8: KEDA scale rule Original deck

    Scaler metadata key for Service Bus vs Storage queue? Polling interval, cool-down, formula, replica defaults?

  9. Card 9: Kubernetes Service types Original deck

    List them. Which is the default? Is Ingress one of them?

  10. Card 10: Rolling update defaults Original deck

    maxSurge and maxUnavailable defaults - for a Deployment, a DaemonSet, and an AKS node-pool upgrade.

  11. Card 11: Kubernetes probes Original deck

    What does each probe failure actually do? What makes a healthy pod unreachable?

  12. Card 12: Container Apps revisions Original deck

    What are the revision modes? What is required before traffic weights work?

  13. Card 13: Container Apps scale setup

    What goes into a custom scale rule? When is min replicas 1 better than 0? Which rules can never reach zero?

  14. Card 14: Container Apps secrets

    Where are Container Apps secrets scoped? How do you point one at Key Vault, and how does the container read it?

  15. Card 15: AKS manifests and rollouts

    What format is a Kubernetes manifest and which command applies it? Order the steps of a zero-downtime image update.

  16. Card 16: AKS autoscalers

    HPA, VPA, cluster autoscaler, KEDA - what does each one scale? What must containers define before HPA works?

  17. Card 17: AKS pod troubleshooting

    A pod keeps restarting or its Service returns nothing. What do you check first, and what does each status mean?

Domain 2 Develop AI solutions by using Azure data management services 20 cards

  1. Card 18: Cosmos consistency levels Original deck

    All five, strongest to weakest. Which is default? What are the bounded-staleness minimums?

  2. Card 19: Cosmos vector indexes Original deck

    The three types, their max dimensions, and the threshold before an index is used.

  3. Card 20: Cosmos RU and throttling Original deck

    What costs 1 RU? Autoscale range? What happens when you exceed provisioned throughput?

  4. Card 21: Cosmos change feed Original deck

    The two modes, their limits, and what the processor requires.

  5. Card 22: Cosmos indexing policy Original deck

    What are the defaults, and how do you cut write RU cost?

  6. Card 23: Cosmos SDK object model

    Name the three SDK objects in the order you get them. Which one holds shared throughput? Which one runs item CRUD?

  7. Card 24: Cosmos vector search

    How do you store an embedding, keep vector writes cheap, and return the 5 nearest items? Which index does NOT help?

  8. Card 25: Change feed scale-out

    What are the four processor parts? How do several instances share the work, and what do Python apps use instead?

  9. Card 26: pgvector operators and indexes Original deck

    The three distance operators, their operator classes, and the dimension limits.

  10. Card 27: PgBouncer on Azure PostgreSQL Original deck

    Ports, how to enable it, the pool mode default, and the tier restriction.

  11. Card 28: PostgreSQL app connections

    How should app code authenticate, reuse connections, survive traffic spikes, and pass user input? Which settings are NOT the fix?

  12. Card 29: pgvector filtered retrieval

    Which column types? In what order do you build it? Write the filtered top-5 query. What cuts CPU when filters narrow the rows?

  13. Card 30: PostgreSQL vector resources

    Match each symptom to a resource. Which tier has the most memory per vCore? In what order do you prove a scale-up worked?

  14. Card 31: Redis eviction policies Original deck

    All eight. What do the volatile ones ignore? What is Azure's default?

  15. Card 32: Expiration vs eviction Original deck

    Different triggers, different settings. How do you tell which a question is about?

  16. Card 33: Redis ports by tier Original deck

    Which port belongs to which service and tier?

  17. Card 34: RediSearch vector search Original deck

    Field types, algorithms, distance metrics, and the create-time constraints.

  18. Card 35: Redis cache freshness

    Which tool gives a fixed lifetime, which keeps data fresh after a source change, and which cuts repeat database reads? What does NOT affect freshness?

  19. Card 36: Keeping hot keys in Redis

    Frequently read keys should stay and rarely read keys should go. When is the answer EXPIRE, and when an eviction policy? What does PERSIST do?

  20. Card 37: Redis vector storage basics

    How do you write an embedding to a hash, and why can a search return nothing? What host, port and TLS setting does Azure Managed Redis use?

Domain 3 Connect to and consume Azure services 14 cards

  1. Card 38: Event Grid retry schedule Original deck

    What is the actual back-off schedule, and what counts as success?

  2. Card 39: Event Grid permanent failures Original deck

    Which status codes are never retried? Retry-policy limits? Dead-letter facts?

  3. Card 40: Event Grid filtering Original deck

    The three filter kinds, and which can see inside the payload?

  4. Card 41: Service Bus settlement Original deck

    The four actions. Which increments the delivery count?

  5. Card 42: Service Bus numbers Original deck

    Lock duration, max delivery count, duplicate detection window, and the FIFO rule.

  6. Card 43: Event Grid topics and setup

    Setup order for publishing your own events? Custom vs system vs partner topic vs domain? What does the handshake prove?

  7. Card 44: Service Bus queues vs topics

    Which entities fan one message out so every consumer gets a copy? Which Service Bus features are NOT delivery entities?

  8. Card 45: ServiceBusProcessor setup

    The four setup steps in order? What happens if a handler is missing, or the message handler throws?

  9. Card 46: Functions triggers and retries Original deck

    Queue retry behavior, timer format, blob trigger latency, Cosmos trigger requirements.

  10. Card 47: HttpTrigger attribute Original deck

    The authorization levels, the default, and how the method list behaves.

  11. Card 48: Choosing a Functions trigger

    Which trigger for an immediate reply, background work, a schedule, new blobs, changed documents? Why not do slow work in HTTP?

  12. Card 49: Triggers and bindings

    How many triggers per function? What brings blob contents in without SDK code? Which bindings does each service offer?

  13. Card 50: Function app configuration

    Where do per-environment values live? What is local.settings.json for? How does a Key Vault reference work and rotate?

  14. Card 51: Secure function deployment

    Order the steps for a secure function app release. Which option deploys both infrastructure and code from source control?

Domain 4 Secure, monitor, and troubleshoot Azure solutions 17 cards

  1. Card 52: Key Vault data-plane roles Original deck

    Narrowest to broadest. Which one reads secret values? Which is a trap?

  2. Card 53: Key Vault versions and rotation Original deck

    What makes rotation work automatically? What is the rotation mechanism?

  3. Card 54: DefaultAzureCredential chain Original deck

    The documented order in the Python SDK.

  4. Card 55: App Configuration Original deck

    What identifies a key-value? Refresh defaults? Roles? What belongs where?

  5. Card 56: Key Vault access for an app

    Put the setup steps in order. Which scope? Which permission sounds right but does NOT read a secret?

  6. Card 57: Where secrets belong

    Which store holds a credential, and what does it add? Why not App Configuration or Kubernetes secrets?

  7. Card 58: App Configuration Python SDK

    Which call reads one setting? Which lists many? What gives one key a value per environment, and what does NOT?

  8. Card 59: Dynamic configuration refresh

    How does an app pick up changed settings with no restart and no remote call per request? What is the Python pattern?

  9. Card 60: KQL operator shapes Original deck

    What does each operator do to the shape of the data? What is the performance order?

  10. Card 61: The two telemetry schemas Original deck

    App Insights vs Log Analytics table and column names. What unit is duration?

  11. Card 62: KQL time windows and grouping

    Write "the last 30 minutes". How many rows does summarize return? Which join kind do you get by default?

  12. Card 63: Azure Monitor query tools

    One-off investigation of app logs from several services: which tool? What do the activity log and workbooks give you instead?

  13. Card 64: OpenTelemetry pipeline Original deck

    The four parts and what each one owns. What breaks if one is missing?

  14. Card 65: Tracing setup traps

    Which actions make distributed traces appear and correlate? Which look related but do NOT create spans?

  15. Card 66: OpenTelemetry samplers Original deck

    The real spec names, the Python names, and the environment-variable values.

  16. Card 67: Picking a trace sampler

    Which sampler records every span? Which keeps 10 percent? What is NOT a sampler, and how does Azure Monitor sample?

  17. Card 68: Trace context propagation Original deck

    What carries the trace across a service boundary, and what has to be installed?