AI Bot
AI-103: Deployment, Monitoring, and Security Flashcards
Microsoft Azure AI App and Agent Developer Associate AI-103 Flashcards
Study our AI-103: Deployment, Monitoring, and Security flashcards for the Microsoft Azure AI App and Agent Developer Associate AI-103 exam with 55+ flashcards. View as flashcards, a searchable table, or as a fun matching game.
| Front | Back |
| A B testing for models | Compare model variants by routing user traffic |
| ACI purpose | Serverless containers for single container workloads |
| AKS purpose | Managed Kubernetes service for container orchestration |
| Alerting strategy | Use thresholds anomaly detection and escalation policies |
| Application Insights purpose | End to end monitoring tracing and telemetry for apps |
| Autoscaling strategies | Horizontal scaling vertical scaling and instance scaling |
| Azure AD role | Central identity provider for users services and apps |
| Azure container options | AKS ACI App Service for Containers |
| Azure Cost Management purpose | Monitor analyze and optimize cloud spend |
| Azure DevOps pipelines role | CI CD orchestration and build agent management |
| Azure Functions use case | Serverless compute for event driven workloads |
| Azure Monitor role | Collect metrics logs and alerts for Azure resources |
| Backup and recovery for models | Regular snapshot artifacts and store in durable storage |
| Blue green deployment definition | Run parallel stable and new environments then switch |
| Canary deployment definition | Gradually shift traffic to new model version |
| CI CD for models | Automate build test and deployment of model artifacts |
| CI CD testing types | Unit tests integration tests data validation and model tests |
| Compliance on Azure | Use Azure Policy Blueprints and compliance manager |
| Cost optimization tactics | Right size instances use spot instances and autoscaling |
| Data anonymization technique | Remove or obfuscate personal identifiers before use |
| Data encryption at rest | Use AES based encryption managed by Azure services |
| Data encryption in transit | Use TLS for secure network communications |
| Dependency management for AI | Pin package versions use containers and reproducible builds |
| Differential privacy concept | Add noise to outputs to protect individual data |
| Distributed tracing use | Track requests across services to diagnose latency |
| Explainability requirement | Provide model explanations for audit and compliance |
| GitOps for ML | Use Git as single source of truth for pipelines and deployments |
| Identity principle of least privilege | Grant minimal access required for tasks |
| Immutable infrastructure benefit | Replace not modify for predictable deployments |
| Incident response steps | Detect contain mitigate root cause and recover |
| Key Vault purpose | Securely store and manage secrets keys and certificates |
| Logging best practice | Structured logs centralized and correlated with traces |
| Managed identities benefit | Securely assign Azure resource identities for auth |
| MLOps pipeline components | Data training validation model registry deployment |
| Model drift detection | Monitor performance and input distributions for degradation |
| Model encryption needs | Encrypt sensitive model artifacts and feature data |
| Model provenance importance | Track data code and training environment for reproducibility |
| Model registry purpose | Store version metadata and artifacts for models |
| Monitoring key metrics | Latency throughput error rate and resource utilization |
| Network isolation methods | Use VNets subnets NSGs and private endpoints |
| Postmortem focus | Document timeline causes impact and preventative actions |
| Private endpoints advantage | Securely access PaaS services over private network |
| RBAC vs ABAC | RBAC uses roles ABAC uses attributes for fine grained control |
| Retraining triggers | Performance drop data distribution change or business update |
| Runbook contents | Step by step remediation procedures and contacts |
| Runtime protection tools | Use runtime application self protection and host hardening |
| Scaling stateful models | Use session affinity sticky sessions or external state store |
| Scaling stateless models | Scale out by adding replicas behind load balancer |
| Secret rotation practice | Regularly rotate credentials and automate updates |
| Secure data pipeline elements | Ingress validation encryption access controls and auditing |
| Secure model deployment patterns | Use model signing network isolation and auth |
| Shadow deployment pattern | Run new model in parallel without serving decisions |
| Spot instances tradeoff | Lower cost but potential preemption risk |
| Supply chain security for ML | Scan dependencies sign artifacts and verify provenance |
| When to use managed services | Reduced ops and built in scaling |
About the Flashcards
Flashcards for the Microsoft Azure AI App and Agent Developer Associate exam help students review terminology, Azure platform choices, and service trade-offs used when deploying models and applications. The deck defines container options (AKS, ACI, App Service), serverless compute like Azure Functions, and practical guidance on when to use managed services.
It also covers MLOps pipelines, CI/CD practices, and model lifecycle essentials such as model registry, testing types, and reproducible builds. Deployment patterns (canary, blue-green, A/B, shadow), scaling strategies, monitoring, cost optimization, security controls, identity and Key Vault, incident response, and backup/recovery are included for targeted review.
Topics covered in this flashcard deck:
- Azure container options
- MLOps pipeline components
- Deployment patterns
- Scaling and autoscaling
- Monitoring and logging
- Security, identity, encryption