Agentic AI on AWS
This job-oriented course moves from Agentic AI fundamentals to production deployment. You will work with Amazon Bedrock, foundation models, Bedrock Agents, Knowledge Bases, RAG, vector databases, Lambda tools, memory, multi-agent workflows, AWS AI services, security, governance, Python, frameworks, observability, and MLOps.
Curriculum Areas
- Module 1 - Introduction to Agentic AI
- Module 2 - AWS AI & Generative AI Fundamentals
- Module 3 - Amazon Bedrock
- Module 4 - Building AI Agents with Amazon Bedrock
- Module 5 - Prompt Engineering
- Module 6 - RAG on AWS
- Module 7 - Vector Databases & Knowledge Bases
- Module 8 - Agent Tools & Action Groups
- Module 9 - AWS Lambda + Agentic AI
- Module 10 - Agent Memory & Context
- Module 11 - Multi-Agent Systems
- Module 12 - Agentic Workflows
- Module 13 - AWS AI Services Integration
- Module 14 - Agentic AI with AWS Data
- Module 15 - Security for Agentic AI
- Module 16 - Responsible AI & Governance
- Module 17 - Agentic AI with Python
- Module 18 - Agentic AI Frameworks
- Module 19 - Observability & Evaluation
- Module 20 - Production Deployment / MLOps
- Hands-On Real-Time Projects
Agentic AI on AWS Curriculum
- What is Agentic AI?
- Generative AI vs Agentic AI
- AI Agents vs Chatbots
- Agent architecture
- LLM reasoning and decision-making
- Tools, actions, memory and planning
- Single-agent vs multi-agent systems
- Real-world enterprise use cases
- AWS AI/ML ecosystem
- Generative AI on AWS
- Amazon Bedrock overview
- Foundation Models (FMs)
- Model selection
- Tokens, context windows and inference
- Temperature and other model parameters
- AWS account, IAM and basic AWS services
- Amazon Bedrock architecture
- Foundation models in Bedrock
- Model access and configuration
- Bedrock APIs
- Bedrock playground
- Inference
- Prompt engineering
- Model evaluation
- Guardrails
- Cost and performance considerations
- What is an AI Agent?
- Bedrock Agents architecture
- Agent instructions
- Agent orchestration
- Action Groups
- Knowledge Bases
- Agent session management
- Agent invocation
- Testing and debugging
- Deploying production agents
- Prompt engineering fundamentals
- Zero-shot and few-shot prompting
- Role-based prompting
- Structured prompts
- Chain-of-thought concepts
- Context engineering
- Prompt templates
- Dynamic prompts
- Output formatting
- Prompt optimization
- What is Retrieval-Augmented Generation?
- RAG architecture
- Document ingestion
- Chunking
- Embeddings
- Vector search
- Retrieval
- Context augmentation
- Amazon Bedrock Knowledge Bases
- RAG evaluation
- Reducing hallucinations
- Vector database fundamentals
- Amazon OpenSearch Serverless
- Amazon Aurora PostgreSQL with vector capabilities
- Embedding models
- Similarity search
- Metadata filtering
- Knowledge Base configuration
- Connecting enterprise documents to agents
- Why agents need tools
- Tool/function calling
- Action Groups
- Lambda-based actions
- API-based actions
- OpenAPI schemas
- Connecting agents to databases
- Calling external services
- Error handling
- Lambda fundamentals
- Creating agent tools with Lambda
- Event-driven AI workflows
- Lambda + Bedrock integration
- API Gateway integration
- Calling external APIs
- Exception handling
- Monitoring Lambda-based agents
- Short-term memory
- Long-term memory concepts
- Conversation history
- Session management
- User context
- Personalization
- State management
- Managing large context windows
- Multi-agent architecture
- Supervisor agents
- Specialist agents
- Agent collaboration
- Agent-to-agent communication
- Task delegation
- Parallel agent execution
- Multi-agent workflows
- Enterprise use cases
- Sequential workflows
- Parallel workflows
- Conditional workflows
- Human-in-the-loop
- Planning and execution
- Tool selection
- Workflow orchestration
- Retry and recovery mechanisms
- Amazon Comprehend
- Amazon Textract
- Amazon Transcribe
- Amazon Polly
- Amazon Rekognition
- Amazon Translate
- Integrating AI services into agent workflows
- Building multimodal applications
- Amazon S3
- Amazon DynamoDB
- Amazon RDS
- Amazon Aurora
- OpenSearch
- Data ingestion pipelines
- Enterprise data grounding
- Secure data access
- IAM for AI applications
- Roles and policies
- Least-privilege access
- Encryption
- Secrets management
- Network security
- Data privacy
- Guardrails
- Prompt injection
- Data leakage prevention
- Secure tool execution
- AI safety
- Hallucination management
- Bias and fairness
- Content filtering
- Guardrails
- Explainability
- Auditability
- Human oversight
- AI governance policies
- Python fundamentals for AI applications
- AWS SDK for Python (Boto3)
- Calling Bedrock APIs
- Working with models
- Prompt management
- Tool calling
- Building AI agents with Python
- API integration
- Error handling
- Agent frameworks overview
- LangChain
- LangGraph
- AWS integrations
- State management
- Tool calling
- Agent orchestration
- Building custom agent workflows
- Agent performance monitoring
- CloudWatch
- Logging
- Tracing
- Agent evaluation
- Response quality
- Latency monitoring
- Token usage
- Cost monitoring
- Troubleshooting production agents
- Development to production
- CI/CD for AI applications
- Infrastructure as Code
- AWS CloudFormation / CDK concepts
- Containerization
- API deployment
- Scalability
- High availability
- Production monitoring
Project 1 - AWS Customer Support Agent
- Answers customer questions
- Searches company documents
- Uses RAG
- Creates support tickets
- Calls external APIs
- Escalates to a human
Project 2 - Enterprise RAG Agent
- Upload enterprise documents to S3
- Create embeddings
- Build Knowledge Base
- Implement vector search
- Connect Bedrock Agent
- Implement secure document retrieval
Project 3 - AI Sales Agent
- Retrieve customer information
- Qualify leads
- Generate personalized responses
- Update CRM data
- Schedule follow-ups
- Use external APIs