Generative AI Online Training Course Content
๐ ๏ธ Basics & Operations
- ๐ข OPERATORS
- ๐ MATH LIBRARY
- ๐ฆ VARIABLES
- ๐ DATA TYPES
- ๐ TYPECASTING
- โ BOOLEANS
๐งต String Manipulation
- ๐ค STRINGS
- โจ SPECIAL CHARACTERS IN A STRING
- โ๏ธ SPLIT & STRIP A STRING
๐๏ธ Data Structures
- ๐ LISTS
- ๐ DICTIONARY
- ๐ก SETS
- ๐ TUPLES
โ๏ธ Logic & Functions
- ๐ฆ IF, WHILE, FOR, etc.
- ๐๏ธ FUNCTIONS
- ฮป LAMBDA FUNCTIONS
๐๏ธ Fundamentals
- ๐ Population Vs Sample
- ๐ Types of Variables
- ๐๏ธ Branches of Statistics
- ๐ Statistics Library
๐ Descriptive Statistics
- ๐ Central Tendency
- ๐ Measures of Dispersion
- ๐ Variance Vs. Standard Deviation
- ๐ Distribution
๐ Relationships & Probability
- ๐ฏ Z-Score
- ๐ Correlation
- ๐ Covariance
- ๐ฒ Probability
- ๐ค Models
๐งช Inferential Statistics
- ๐ฌ Sampling Methods
- ๐งฌ Sampling Techniques
- ๐ Central Limit Theorem
- ๐ Confidence Interval
- โ๏ธ Hypothesis Testing
- ๐ Statistical Significance
๐ Module 3: Maths
- ๐ข Algebra Basics
- ๐ Calculus
- ๐งฑ Matrices
๐๏ธ Fundamentals & Data
- ๐ท๏ธ Labelled Vs Unlabeled Data
- ๐งฉ Types of ML Algorithms
- ๐ฎ How ML Predict things
- โ๏ธ Train Test Split
๐ Text Preprocessing (NLP)
- ๐ข Count Vectorizer
- โจ Special & Numerical Characters
- ๐ Remove HTML Tags from Text
- ๐ Remove Stop words from Text
- ๐ฑ Stemming
๐ Regression & Analysis
- ๐ Simple Linear Regression
- ๐ Multiple Linear Regression
- ๐ Polynomial Regression
- โณ Univariate Timeseries Analysis
- ๐ Loss Functions
๐งช Classification & Clustering
- โ๏ธ Classification Algorithms
- ๐๏ธ Clustering Algorithms
- ๐ ROC & AUROC
๐ ๏ธ Optimization & Engineering
- ๐ Difference between fit and fit_transform Methods
- ๐ Dimensionality Reduction
- ๐งฌ Principal Component Analysis (PCA)
- โ๏ธ HyperParameters
- ๐ Noise
- โ๏ธ Penalty
โ๏ธ Evaluation Metrics
- ๐ฏ Accuracy โ MAE, MSE, RMSE & Variance Score
- ๐ $r^2$ score
๐ค Ensemble Methods
- ๐ Average Ensemble Method
- โ๏ธ Weighted Ensemble Method
- ๐ Conditional Ensemble Method
- ๐ Bagging Ensemble Method
- ๐ Boosting Ensemble Method
๐๏ธ Templates & Structure
- ๐ Project Skeletons
- ๐ ML Template Regression
- ๐ ML Template Classification
๐งฑ Neural Network Foundations
- ๐ธ๏ธ Artificial Neural Network
- ๐๏ธ Deep Neural Network
- ๐งฌ Neuron
- โ๏ธ Weight
- ๐ Bias
๐๏ธ Architecture Layers
- ๐ฅ Input Layers
- ๐ป Hidden Layers
- ๐ค Output Layer
๐ Sequence & Image Models
- ๐ RNN (Recurrent Neural Networks)
- ๐ง LSTM (Long Short-Term Memory)
- ๐ผ๏ธ CNN (Convolutional Neural Networks)
๐จ Generative & Vision AI
- ๐ญ Generative Adversarial Network (GAN)
- ๐ Diffusion Models
- ๐ฏ YOLO Object Detection
- ๐๏ธ Vision Transformer (ViT)
๐๏ธ Orchestration & Goal
- ๐ฏ Master the low-code orchestration layer
- โ๏ธ Connect AWS services, LLMs, and external tools
๐๏ธ n8n Architecture
- ๐งฉ Nodes, Workflows, and JSON data flow
- ๐ HOSTING: Running n8n on AWS EC2 (Docker) vs Cloud
- โก Webhooks: The bridge between AWS Lambda/API Gateway and n8n
๐ ๏ธ Core Integration Patterns
- ๐ HTTP Request Node: Connecting to Amazon Bedrock API
- ๐ป Code Node: Writing custom Python/JS logic within workflows
๐ Data Manipulation
- ๐ Transforming JSON for LLM context
- ๐ฆ Handling binary data (images/files) from S3
๐ฏ Core Goal
- ๐ ๏ธ Mastery: Build tool-using agents visually without complex codebases.
๐ง AI Agent Nodes
- โ๏ธ Agent Logic: Configuring ReAct vs. Chain-of-Thought reasoning.
- โ๏ธ Model Connection: Integrating Amazon Bedrock (Claude 3.5 Sonnet).
๐ ๏ธ Tool Calling & MCP
- ๐ง Custom Tools: Creating functional tools via n8n workflows.
- ๐ MCP Integration: Connecting to local/remote Model Context Protocol servers.
- ๐ External Tools: Integrating Tavily Search, Gmail, and Slack.
๐พ Memory Management
- ๐ง Short-Term: WindowBufferMemory for immediate context.
- ๐ Long-Term: VectorStoreMemory using Amazon OpenSearch.
๐ฏ Goal: Orchestrate complex, multi-step agent systems.
๐ฐ Master-Worker Architecture
- ๐ Calling Sub-workflows: Utilizing specialized agents for modular tasks.
- ๐ค Orchestrating a “Swarm”: Coordinating roles like Researcher, Writer, and Reviewer.
๐ฆ Advanced Logic
- ๐ Switch/Router Nodes: Implementing conditional execution paths.
- ๐ Looping over Datasets: Efficient batch processing of information.
๐ค Human-in-the-Loop
- โณ Wait Nodes: Pausing for email or form approvals before execution.
๐ฏ Goal
- ๐ Build multi-modal agents that can speak and listen.
๐ Voice Integration
- ๐ Integrating VAPI or Retell AI for telephony.
- โก Triggering n8n workflows from voice inputs.
โก Real-time Interaction
- ๐ Handling low-latency responses with Groq or Bedrock instantaneous models.
- ๐ฑ Multi-modal triggers (Voice -> Action -> SMS).
๐ผ Module 10: The AI Automation Agency (AAA) Business Model
๐ฏ Goal
- ๐ฐ Structure and monetize your AI services.
๐ Business Fundamentals
- ๐ข The Agency Model: Retainers vs Project-based pricing.
- ๐ฏ Selecting a niche (Real Estate, Healthcare, E-commerce).
๐ Delivery Systems
- ๐ Client Onboarding & Staging environments.
- ๐ Documentation as a deliverable.
๐ก๏ธ Reliability & Monitoring
- ๐จ Universal Error Triggers (Sentry/Slack alerts).
- ๐ Uptime monitoring and auto-scaling on AWS.
๐ Module 11: Agentic AI Essentials on AWS
๐ฏ Goal
- ๐ข Understand the shift from Generative AI to Agentic AI and get introduced to the AWS tools that enable it.
๐ค Agentic AI Introduction
- โ๏ธ AI Agents vs. Agentic AI vs. Traditional AI.
- ๐งฉ Core Building Blocks: Brain, Tools, Planning, Memory.
๐ ๏ธ The AWS Agentic Stack
- โ๏ธ Amazon Bedrock: The foundation model layer (Claude 3.5 Sonnet, Titan, Llama 3).
- ๐ Amazon Q: The built-in assistant for AWS development.
- โก AWS Lambda: The compute layer for agent tools.
๐ก๏ธ Ethical and Responsible AI on AWS
- ๐ง Introduction to Guardrails for Amazon Bedrock (PII redaction, content filtering).
- ๐ Best practices for secure agent design (IAM roles, Least Privilege).
๐๏ธ Module 12: Agentic AI Architectures & Serverless Design Patterns
๐ฏ Goal
- ๐ Learn how to design scalable agent architectures using AWS serverless patterns.
๐๏ธ Architecture Types
- ๐ Router, Planner-Executor, Supervisor-Worker.
- โก AWS Pattern: Event-driven agents using Amazon EventBridge.
๐งฉ Core Modules
- ๐๏ธ Perception (Bedrock)
- ๐ง Cognition (Bedrock)
- ๐ ๏ธ Action (Lambda)
- ๐ก๏ธ Security (IAM/Guardrails)
โ๏ธ Design Considerations
- โฑ๏ธ Latency: Provisioned Throughput for Bedrock.
- ๐ฐ Cost: Token management and Bedrock pricing models.
- ๐ก๏ธ Reliability: Retry mechanisms with AWS Step Functions.
๐ฏ Goal
- ๐ป Build your first agents using code-first approaches with LangChain on AWS.
๐ ๏ธ Working with Amazon Bedrock API
- ๐ก Invoking models using
boto3andlangchain-aws. - ๐ Streaming responses.
๐พ Memory & State
- ๐๏ธ Storing conversation history in Amazon DynamoDB.
๐ง Tool Use
- ๐ ๏ธ Defining functional tools (Function Calling) with Claude 3.5 Sonnet.
- โ๏ธ Connecting agents to AWS APIs (e.g., EC2, S3).
๐ Deployment Patterns
- ๐ฆ Deploying LangChain agents as AWS Lambda functions or Fargate containers.
๐ฏ Goal
- ๐ Master stateful, multi-step agent workflows using LangGraph extended with AWS persistence.
๐ข LangGraph Basics
- ๐ Nodes, Edges, and State Schema.
๐พ State Persistence on AWS
- ๐๏ธ Implementing a custom Checkpointer using Amazon DynamoDB or Aurora PostgreSQL.
๐ค Human-in-the-Loop
- ๐ Implementing approval gates (interruptions) in LangGraph.
- ๐ Storing approval state in DynamoDB.
๐ Deployment
- โ๏ธ Host the graph traversal loop on AWS Lambda (or ECS for long-running).
๐ฏ Goal
- ๐ ๏ธ Implement robust Retrieval-Augmented Generation (RAG) systems without managing vector DB infrastructure.
๐ง Knowledge Bases for Amazon Bedrock
- ๐ฅ Ingesting data from S3, Web, Confluence.
- ๐พ Managed embeddings and vector storage (OpenSearch Serverless).
๐ Advanced RAG Strategies
- ๐ Hybrid Search (Text + Vector) in Amazon OpenSearch Serverless.
- ๐ Query rewriting and reranking using Bedrock.
๐ Evaluation
- โ๏ธ Measuring groundedness and relevance using Amazon Bedrock Model Evaluation.
๐ฏ Goal
- ๐บ๏ธ Model complex relationships and reasoning pathways using Graph Databases.
๐ Graph Fundamentals
- ๐ฟ Property Graphs vs. RDF.
- โ๏ธ Introduction to Amazon Neptune (Serverless).
๐๏ธ Modeling for Agents
- ๐ Entity-Relationship modeling for enterprise data.
- ๐ง Ontology-driven design.
๐ Querying
- โก Using Gremlin or OpenCypher for graph traversal.
- ๐ Neptune Analytics for high-speed graph algorithms.
๐ฏ Goal
- ๐ง Combine Graph Databases with LLMs for “GraphRAG” โ retrieving context via relationships, not just similarity.
๐๏ธ GraphRAG Architecture
- ๐ป LLM-driven Cypher/Gremlin generation.
๐ Hybrid Retrieval
- ๐ Vector Search (OpenSearch) + Graph Traversal (Neptune).
๐ Response Patterns
- ๐ “Show evidence + show path” (Explainability).
- ๐งฑ Grounded reasoning using graph relationships.
๐ฏ Goal
- ๐ Understand the emerging standard for connecting AI models to data and tools securely.
๐งฑ MCP Fundamentals
- ๐๏ธ MCP architecture and tool contracts.
- ๐ฅ๏ธ MCP servers and standardized tool interfaces.
๐ก๏ธ Security & Governance
- ๐ Secure tool access (read-only by default).
- ๐ Auditable tool execution.
๐ ๏ธ Implementation
- ๐ Standardizing tool access for agents.
- ๐ Secure tool design.
- ๐งฉ Building extensible agent tooling.
๐ฏ Goal
- โ๏ธ Use the fully managed “Agents for Amazon Bedrock” service to build tool-using agents without managing infrastructure.
๐๏ธ Bedrock Agents Architecture
- โ๏ธ Action Groups: Connecting to Lambda functions via OpenAPI schemas.
- ๐ Knowledge Base integration.
๐ Traceability
- ๐ง Visualizing the “Chain of Thought” in the Bedrock Console.
๐ก๏ธ Security
- ๐ Resource-based policies and execution roles.
๐ฏ Goal
- ๐ Build safe, robust tools and secure the agent’s execution environment.
๐งฑ Guardrails for Amazon Bedrock
- ๐ซ Denying unsafe topics.
- ๐ Redacting PII (Personal Identifiable Information) in real-time.
๐ ๏ธ Tool Safety
- โ Input validation in AWS Lambda.
- ๐ “ReadOnly” vs “ReadWrite” tool modes.
๐ค Human Approval Workflows
- ๐ Using AWS Step Functions to insert a manual approval step before executing high-risk actions.
๐ฏ Goal
- ๐ Monitor, trace, and evaluate agent performance in production.
๐ Observability
- ๐ฐ๏ธ Tracing traces with AWS X-Ray and Amazon CloudWatch.
- ๐ Logging full prompts and completions.
โ๏ธ Evaluation
- ๐ค Running automated evaluation jobs in Amazon Bedrock (Accuracy, Robustness, Toxicity).
- ๐จโโ๏ธ “LLM-as-a-Judge” patterns using Bedrock.
๐ฏ Goal
- ๐จ Build agentic workflows visually.
โก AWS Step Functions
- ๐ผ Orchestrating Bedrock calls in a visual state machine.
- ๐ ๏ธ Handling retries, errors, and parallel processing.
- ๐ก Integration with Amazon EventBridge for event-driven agents.
๐ข Amazon Q Business
- ๐ค Building internal enterprise assistants with no code.
๐ฏ Goal
- ๐ Ensure your agents access data securely and compliantly.
๐ AWS Lake Formation
- ๐ฏ Fine-grained column/row level access control.
๐๏ธ AWS Glue Data Catalog
- ๐ท๏ธ Metadata management for Text-to-SQL agents.
๐ Amazon Athena
- โก Secure query execution for data analysis agents.
๐ฏ Goal
- ๐ Fast-track your dev environment with the new Claude Code CLI tool.
โ๏ธ Initialization & Configuration
- ๐ฅ Installation via npm or bun.
- ๐ Authenticating and managing Access Tokens.
- ๐ก๏ธ Configuring allowed tools and permission scopes (Read/Write/Execute).
๐ง Core Concepts
- ๐ Understanding Sessions & Context Compaction.
- ๐ฆ Sandboxing: Docker-based vs Native execution environments.
- ๐ Undo/Redo actions and Version Control integration.
๐ฏ Goal
- ๐ค Leverage Claude as an autonomous coding partner.
๐ Project Structure
- ๐ Current directory as context.
- ๐ CLAUDE.md: Defining project-specific instructions and style guides.
๐ Plan Mode
- ๐บ๏ธ Breaking down complex features into step-by-step execution plans.
- ๐ Iterative code generation.
๐ ๏ธ Tooling & Subagents
- ๐งฐ Using built-in tools (File edit, Terminal run).
- ๐ Connecting MCP Servers (Model Context Protocol) for database/API access.
- ๐ฅ Creating custom Subagents for specialized tasks.
๐ฏ Goal
- ๐ Extend Claude Code into a fully automated CI/CD and testing loops.
๐ก Skills & Hooks
- ๐ Defining custom Agent Skills (Bash scripts as tools).
- ๐ Using Hooks (pre-command, post-command) for validation.
๐ Testing & Feedback Loops
- ๐งช Providing feedback via automated test outputs.
- ๐ Granting Browser Access for UI validation (Headless browsing).
- โพ๏ธ Running Claude Code in a loop (‘Ralph’ loop) for continuous refactoring.
๐ Overview
- ๐ฏ These projects demonstrate end-to-end competency in the AWS Agentic ecosystem.
๐ง Project 1: Enterprise GraphRAG Agent (Neptune + Bedrock)
- ๐ Scenario: An internal “Corporate Brain” for organizational structure and project queries.
- ๐๏ธ Architecture:
- ๐พ Data: Amazon Neptune (Employee/Project data).
- ๐ง LLM: Claude 3.5 Sonnet via Amazon Bedrock.
- โ๏ธ Logic: LangChain agent on AWS Lambda.
- โจ Key Features:
- ๐ฃ๏ธ Translates natural language to Gremlin/OpenCypher.
- ๐ Performs multi-hop reasoning.
- ๐ Returns graph visualizations (node-link JSON).
๐ก๏ธ Project 2: Serverless Code Reviewer Supervisor (LangGraph on AWS)
- ๐ Scenario: Multi-agent system for PR reviews, fixes, and security audits.
- ๐๏ธ Architecture:
- ๐ผ Orchestration: LangGraph on AWS Lambda.
- ๐พ State: Amazon DynamoDB.
- ๐ค Agents: “Reviewer”, “Security Auditor”, “Refactorer”.
- โจ Key Features:
- ๐ฆ Supervisor node routes work to specialized sub-agents.
- ๐ค Human-in-the-loop: Approval gates for GitHub/CodeCommit.
๐ฅ Project 3: “HR Assistant” Managed Bedrock Agent
- ๐ Scenario: Managed agent for leave requests and policy questions.
- ๐๏ธ Architecture:
- โ๏ธ Platform: Agents for Amazon Bedrock.
- ๐ Knowledge: Bedrock Knowledge Base (S3 Handbook).
- ๐ ๏ธ Action Group: AWS Lambda + DynamoDB.
- โจ Key Features:
- ๐ฌ Conversational context retention.
- ๐ Traceability view in Bedrock Console.
- ๐ซ Guardrails for sensitive info (Salary/PII).
๐ Project 4: Governed Text-to-SQL Data Analyst
- ๐ Scenario: Secure natural language querying of sales data for business users.
- ๐๏ธ Architecture:
- ๐๏ธ Data: S3 + AWS Glue Data Catalog.
- ๐ Query: Amazon Athena.
- ๐ง Agent: Amazon Bedrock (Claude).
- โจ Key Features:
- ๐ท๏ธ Injects Glue Schema into prompts.
- ๐ก๏ธ Lake Formation for column-level security & masking.
- โ Step Functions for SQL approval workflow.
๐ ๏ธ Project 5: Production RAG Pipeline with Eval
- ๐ Scenario: Scalable customer support bot for technical products.
- ๐๏ธ Architecture:
- ๐ฅ Ingestion: Automated S3 โ Bedrock Knowledge Base sync.
- ๐ฅ๏ธ Frontend: Streamlit on EC2 or Fargate.
- โ๏ธ Eval: Scheduled Bedrock Model Evaluation jobs.
- โจ Key Features:
- ๐ Guardrails to block competitor mentions.
- ๐ Custom metric tracking via CloudWatch.
- ๐งช A/B testing for search strategies.
Would you like me to generate a README template or a resource deployment checklist for any of these specific projects?