Practice
92 questions across 5 domains. Includes 20 cross-domain questions.
Select Domain
Select Level
92 questions selected
Drills
Ordering Sequences
RAG Pipeline
- 1 Ingest
- 2 Chunk
- 3 Embed
- 4 Index
- 5 Query
- 6 Retrieve
- 7 Augment prompt
- 8 Generate
- 9 Filter output
Agent Loop
- 1 Receive query
- 2 Plan/reason
- 3 Select tool
- 4 Execute
- 5 Observe result
- 6 Decide (continue/respond)
- 7 Return
Security Layers
- 1 IAM auth
- 2 API Gateway validation
- 3 Guardrails input
- 4 Model invocation
- 5 Guardrails output
- 6 Response encryption
Knowledge Base Setup
- 1 Create S3 bucket
- 2 Upload docs
- 3 Create KB
- 4 Select embeddings model
- 5 Select vector store
- 6 Configure chunking
- 7 Sync data source
- 8 Test retrieval
Deployment Pipeline
- 1 Code
- 2 Build
- 3 Test
- 4 Stage
- 5 Canary
- 6 Production
- 7 Monitor
- 8 Rollback if needed
Prompt Flow Execution
- 1 Input
- 2 Pre-processing node
- 3 Condition check
- 4 Branch A (prompt chain) OR Branch B (direct response)
- 5 Post-processing
- 6 Guardrails
- 7 Output
Evaluation Pipeline
- 1 Define metrics
- 2 Create evaluation dataset
- 3 Select evaluator model
- 4 Run evaluation job
- 5 Analyze results
- 6 Compare models
- 7 Deploy winner
- 8 Monitor in production
Cross-Region Inference Setup
- 1 Enable inference profile
- 2 Update app to use inference profile ARN
- 3 Configure IAM with region conditions
- 4 Test failover
- 5 Monitor with CloudWatch
Matching Exercises
Service -> Data Type
| Trigger / Pattern | AWS Service / Answer |
| Comprehend | Text entities |
| Transcribe | Audio |
| Textract | Documents |
| Rekognition | Images/Video |
| Bedrock Data Automation | Mixed/multimodal |
Problem -> Solution
| Trigger / Pattern | AWS Service / Answer |
| Hallucination | Contextual grounding |
| PII leak | Sensitive info filter (ANONYMIZE) |
| Rate limits | Circuit breaker (Step Functions) |
| Mathematical incorrectness | Automated Reasoning |
| Prompt injection | Prompt Attack filter |
Architecture Pattern -> AWS Service
| Trigger / Pattern | AWS Service / Answer |
| Circuit breaker | Step Functions |
| Model cascading | Lambda routing |
| Semantic cache | Vector DB (ElastiCache/DynamoDB) |
| Prompt governance | Bedrock Prompt Management + CloudTrail |
| Agent deployment | AgentCore Runtime |
Human Review Pattern -> Service
| Trigger / Pattern | AWS Service / Answer |
| Low-confidence routing | A2I (Augmented AI) |
| Evaluation datasets | Ground Truth / Ground Truth Plus |
| Approval workflows | Step Functions Wait-for-Callback |
Guardrails Component -> Verification Type
| Trigger / Pattern | AWS Service / Answer |
| Content filter | Harmful content |
| Contextual grounding | Hallucination |
| Automated Reasoning | Logical correctness |
| PII filter | Data protection |
| Prompt Attack | Jailbreak / injection |
Cost Optimization -> Technique
| Trigger / Pattern | AWS Service / Answer |
| Repeated system prompts | Bedrock Prompt Caching |
| Similar user queries | Semantic caching |
| Mixed complexity queries | Model cascading / Intelligent Prompt Routing |
| Non-real-time bulk processing | Batch inference (50% discount) |
| Idle SageMaker endpoint | Auto-scaling to zero |
External Practice Exams