AI Guardrails & Cybersecurity - Generative AI, MCP Hands On
- Understand the fundamentals of AI Guardrails and their importance in ethical AI development.
- User Input Guardrails : Learn about prompt injections, user input moderations (hate, violence etc) and ways to detect user input violations
- Evaluators - Faithfulness Evaluator(LLM-As-A-Judge), SAS Evaluator, Context Relevance Evaluator and RAGAS Evalauator
Course Content
17 sections • 87 lectures • 6h 35m total length
4.9 (2 reviews)
₹ 2779
6.5 hours on-demand video
Access on mobile and TV
Full lifetime access
Certificate of completion
Lesson 1. Introduction
- Welcome to AI Guardrails
- Course Contents
Lesson 2. General Concepts
- 10,000 Foot view on Language Models
- LLM Inference Parameters
- Vector Embedding
- RAG – Retrieval Augment Generation
Lesson 3. Large Language Models
- Introduction to Large Language Models (LLMs)
- LLM Constraints in Modern AI Applications
- LLM Constraints – Hallucination
- LLM Constraints – Bias and Ethical Concerns
- LLM Constraints – Data Privacy and Security
- LLM Constraints – Output Alignment
- Understanding Constraints in Large Language Models
Lesson 4. AI Guardrails
- Introduction to AI Guardrails
- Applications of AI Guardrails
- Understand AI Guardrails
Lesson 5. User Input Guardrails
- Prompts
- Prompt Injection
- Prompt Guard
- Hands on with prompt-guard model
- Llama Guard Theory
- Llama Guard Prompt
- Hands on with Llama Guard Model
Lesson 6. Multimodel Guardrails
- Llama Guard 3 – Vision Theory
- Hands On with Llama Guard 3 – Vision
Lesson 7. Guardrails on Hallucination
- Hallucination
- Detect Hallucination
- Detect Hallucination with phi3-hallucination-judge model
- Detect Hallucination with hallucination-evaluation-model
Lesson 8. Guardrails on AWS Bedrock
- AWS Bedrock – Guardrails Components
- Introduction to Bedrock
- Bedrock-Guardrail – Hands On
- Multimodal Guardrails- Image
Lesson 9. Scan LLM Vulnerability
- Introduction to Garak
- Garak Probes
- Install Garak
- Detect LLM Vulnerability – Encoding
- XFilteration
- Detect LLM Vulnerability – XFilteration
- Detect LLM Vulnerability – Profanity
Lesson 10. Advanced Web Vulnerability Scanning with AI Agents
- Inroduction to AI Agents
- Agentic Design with Runtime
- Introduction to CrewAI
- Introduction to Penetration Testing
- Hands On Penetration Test using ZapProxy Too
- AI Agent for Cybersecurity/Penetration Testing
- Penetration Test using Tools Call
- Run Cybersecurity Penetration Test with AI Agent
- Run and Evaluate Cybersecurity Penetration Test with AI Agent
Lesson 11. Bonus - Multimodal AI Agent with Tools, Multi-Hop and ReAct Prompt
- Agentic Use Case with Multimodal, Multi-Hop and ReAct Architecture
- ReACT Prompt for AI Agents
- Run the Agent
- Multi Agent with Multi Tools
Lesson 12. Evals
- What is Haystack Framework
- Types of Evaluators
- Evaluators Runtime
- Amazon Bedrock Evaluator – Retriever and Generate(RAG)
- Amazon Bedrock Evaluator – Model as A Judge
Lesson 13. Evaluator with OpenSource Libraries
- Faithfulness (aka Hallucination) Evaluator
- RAG Evaluator
- SAS Evaluator, ContextRelevance Evaluator
Lesson 14. GuardrailsAI Framework
- What is GuardrailsAI
- Working of GuardrailsAI
- GuardrailsAI – RAIL Specifications
- GuardrailsAI – Understanding the Output Component
- GuardrailsAI – Understanding the Validators
- GuardrailsAI – Built-in Validators
- GuardrailsAI – Validator OnFail Policies
- GuardrailsAI – Understanding the Prompt Component
- GuardrailsAI – Understanding the Guard Component
- Example 1 – Extracting Patient Data from Physician Notes
- Example 2 – Detecting Competitor Presence
- Example 3 – Validator Chains – Competitor Analysis & Toxic Language Detection
- Example 4 – Create a Custom Validator
- Understanding GuardrailsAI Framework
- 25 Limitations of GuardrailsAI
Lesson 15. Nemo Guardrails Framework
- Colang
- Example – Execution
- Understanding Nemo Runtime
- Runtime Nemo Multiple LLM Info Log
- Runtime Nemo Single LLM Config
- Runtime Nemo Single LLM Info Log
Lesson 16. Bonus - MCP (Model Context Protocol)
- Evolution of MCP – Current Solutions and their Limitations
- Client Server Architecture
Lesson 17. Bonus - All About MCP
- MCP Architecture
- MCP Server Components
- MCP Transport Types
- MCP Communication
- MCP Flow Diagram – Server, Client and Host communication over Transport Layer
- MCP E2E Flow
