Reference · Printable

AWS Service Cheat Sheet

Every in-scope AWS service for AIP-C01, grouped by function. Use Ctrl/⌘ + P to print as a one-pager for final-day review.

Core GenAI Services

Amazon Bedrock

Fully managed service to access FMs from multiple providers (Anthropic, Meta, Mistral, Amazon, etc.) via API. No infrastructure to manage. The central service for this exam.

Knowledge Bases
Managed RAG. Connect data sources (S3, web crawlers, Confluence), automatic chunk + embed + store + retrieve.
Agents
Managed agent service. Define action groups (OpenAPI + Lambda), connect knowledge bases, add guardrails. Agent autonomously reasons, plans, calls tools.
AgentCore
Infrastructure layer for deploying and scaling agents in production. Compute, networking, lifecycle.
Guardrails
Configurable safety controls. Deny topics · content filter · word filter · PII · grounding · prompt attack. Applied independently to input and output.
Prompt Management
Version, manage, share prompt templates. Parameterized + approval workflows.
Prompt Flows
Visual workflow builder. Chain prompts, conditional logic, pre/post-processing. No-code.
Model Evaluations
Built-in framework for evaluating FM outputs. Automatic + human evaluation. LLM-as-a-Judge supported.
Model Invocation Logs
Detailed logging of all requests and responses. Output to CloudWatch Logs or S3.
Cross-Region Inference
Automatic routing to models in other regions when primary is at capacity or unavailable.
Data Automation
Automated data processing workflows for document extraction and transformation.

Amazon Titan

Amazon's first-party FM family. Text generation, embeddings, image generation. Default embedding choice in Bedrock.

Amazon SageMaker AI

Full ML platform. For AIP-C01 focus on deployment and operations, not training:

Endpoints
Host custom or fine-tuned models with auto-scaling.
Model Registry
Version control and lifecycle management for models.
JumpStart
Pre-built models; one-click deployment of open-source FMs.
Clarify
Bias detection and model explainability (SHAP).
Ground Truth
Human-in-the-loop data labeling.
Data Wrangler
Visual data preparation.
Processing
Run data processing jobs at scale.
Model Monitor
Detect data/model drift in production.
Neo
Optimize models for edge deployment.
Unified Studio
Unified development environment for ML + GenAI.

Agent Frameworks

Strands Agents

AWS open-source agent framework. Build agents with custom tools, memory, state management. Implements ReAct and other patterns. Run on your own compute.

AWS Agent Squad

Multi-agent orchestration. Coordinate specialized agents working together on complex tasks. Supervisor + workers.

AI Application Services

Amazon Q Business
Enterprise AI assistant. Connects to S3, SharePoint, Confluence, Salesforce. Answers questions over org data. Q Business Apps for no-code custom apps.
Amazon Q Developer
AI coding assistant. Code gen, refactoring, debugging, security scanning. In-IDE and CLI.
Amazon Comprehend
NLP service. Entity recognition, sentiment, key phrases, language detection, PII detection. Used for pre-processing.
Amazon Kendra
Intelligent enterprise search powered by ML. Can serve as retrieval source in RAG.
Amazon Lex
Conversational AI. Build chatbots and voice bots.
Amazon Textract
Extract text, tables, forms from PDFs/images. Feeds FM pipelines.
Amazon Transcribe
Speech-to-text. Convert audio/video to text for FM processing.
Amazon Rekognition
Image and video analysis. Object detection, content moderation.
Amazon Augmented AI (A2I)
Human review workflows. Route low-confidence FM predictions to humans.

Compute & Orchestration

AWS Lambda
Serverless compute. Key for data processing, tool implementations, simple MCP servers. Pay per invocation.
AWS Step Functions
Serverless orchestration. Multi-step workflows with error handling, retries, parallel execution. Essential for agent workflows, RAG pipelines, CI/CD, human-in-the-loop.
Amazon EventBridge
Serverless event bus. Event-driven GenAI architectures.
Amazon ECS / EKS / Fargate
Container orchestration. Complex MCP servers, custom model serving, stateful agent components.
Amazon API Gateway
Managed API. Rate limiting, auth, request transform, caching, WebSocket support. Front-end for GenAI APIs.
AWS AppConfig
Feature flags and configuration. Switch models, toggle guardrails, adjust parameters without redeployment.
AWS Outposts
AWS infrastructure on-premises. Data residency / compliance scenarios.
AWS Wavelength
Edge deployments at 5G locations. Ultra-low latency FM inference.

Data & Storage

AWS Glue
Data Quality (validation rules) · Data Catalog (metadata) · Data Lineage (provenance tracking).
Amazon OpenSearch
Search and analytics. k-NN plugin for vector search, Neural plugin for Bedrock integration. Key vector store.
Amazon DynamoDB
NoSQL. Conversation history, session state, metadata storage, agent memory. Streams for change data capture.
Amazon Aurora
Relational. pgvector extension for vector search when you need SQL + vectors.
Amazon S3
Object storage. Source docs for RAG, prompt templates, model artifacts. Lifecycle policies for retention.
Amazon Kinesis
Real-time data streaming. Stream data into FM pipelines.
Amazon RDS
Managed relational databases. Combined with S3 for hybrid document storage.

Security & Identity

IAM
Policies control who invokes which models, accesses which data. Least privilege is critical.
AWS KMS
Key Management. Encrypt data at rest and in transit. Used by Bedrock and SageMaker.
Amazon Cognito
User authentication and authorization. Identity federation for GenAI apps.
Amazon Macie
Discover and classify sensitive data in S3. Audit data before feeding to FMs.
AWS Secrets Manager
Store and rotate API keys, DB credentials.
AWS WAF
Web Application Firewall. Protect API Gateway endpoints.
AWS PrivateLink / VPC Endpoints
Keep Bedrock traffic off the public internet. Required for many compliance scenarios.
IAM Identity Center
SSO for AWS. Federated access to GenAI resources.

Monitoring & Observability

Amazon CloudWatch
Metrics, alarms, dashboards. Track token usage, latency, error rates, costs.
CloudWatch Logs / Logs Insights
Log storage and analysis. Query prompts and responses at scale.
AWS X-Ray
Distributed tracing. Trace requests across services to identify latency bottlenecks.
AWS CloudTrail
API audit logging. Who called which Bedrock/SageMaker API, when, with what parameters.
AWS Cost Explorer
Track GenAI spending trends.
AWS Cost Anomaly Detection
Detect unexpected cost spikes. Alert on anomalies.
Amazon Managed Grafana
Unified dashboards across AWS services.

Developer Tools

AWS Amplify
Frontend framework. Build web/mobile apps with GenAI backends. Hosting, auth, APIs.
AWS CodePipeline
CI/CD orchestration.
AWS CodeBuild
Build service for CI/CD.
AWS CodeDeploy
Deployment automation.
AWS CDK / CloudFormation
Infrastructure as Code. Define GenAI infrastructure programmatically.
AWS CodeArtifact
Artifact repository for build dependencies.
AWS Tools & SDKs
Language-specific SDKs for Bedrock, SageMaker, etc.
Pro tip This page is print-optimized. Ctrl/⌘ + P gives you a black-on-white printable version for quick reference on exam day (if your testing environment allows it — typically not, but useful for last-day review).
↑