NCP-AAI Agentic AI Professional Study Plan¶

8-Week Intensive Study Schedule¶

Phase 1: Foundation Building (Weeks 1-2)¶

Week 1: Agentic AI Fundamentals¶

Focus: Core agent concepts and patterns

Day 1-2: Agent Architecture Overview¶

• Study the concept of AI agents - autonomy, tool use, reasoning
• Learn ReAct pattern - reasoning and acting interleaved
• Understand Plan-and-Execute pattern for complex tasks
• Study Reflexion pattern for self-improving agents
• Reference: NVIDIA Agentic AI Blog

Day 3-4: Memory and State¶

• Study short-term memory (conversation context)
• Learn long-term memory with vector databases
• Understand episodic memory for learning from experience
• Practice implementing agent state management
• Lab: Build a simple ReAct agent with memory

Day 5-7: Function Calling Basics¶

• Study function calling schema definition (JSON Schema)
• Learn parameter extraction from natural language
• Understand tool selection and routing
• Practice defining tools with clear schemas
• Reference: NIM Function Calling

Week 2: NVIDIA Stack for Agents¶

Focus: NIM and NeMo Guardrails setup

Day 1-2: NVIDIA NIM for Agents¶

• Deploy NIM with a function-calling capable model
• Explore NIM OpenAI-compatible API
• Test function calling through NIM API
• Understand NIM model selection for agentic workloads
• Reference: NIM Documentation

Day 3-4: NeMo Guardrails Introduction¶

• Install NeMo Guardrails framework
• Study Colang language basics
• Implement simple input and output rails
• Understand topical guardrails
• Reference: NeMo Guardrails Docs

Day 5-7: Agent Framework Integration¶

• Integrate NIM with LangChain agent framework
• Build a basic tool-calling agent with NIM backend
• Add NeMo Guardrails to the agent
• Test agent behavior with various inputs
• Lab: Build an end-to-end agent with NIM + Guardrails

Phase 2: Advanced Topics (Weeks 3-5)¶

Week 3: Advanced Tool Use and Function Calling¶

Day 1-2: Tool Orchestration¶

• Study sequential vs parallel tool execution
• Learn pipeline patterns for tool chains
• Implement error handling and retry strategies
• Practice conditional tool selection
• Lab: Build agent with multiple tools and parallel execution

Day 3-4: Custom Tool Development¶

• Design tool schemas with clear descriptions
• Build custom tools for database queries, API calls
• Implement tool validation and error handling
• Test tools independently before agent integration

Day 5-7: Complex Tool Scenarios¶

• Multi-step tool chains with data transformation
• Tool fallback strategies when primary tool fails
• Dynamic tool registration and discovery
• Rate limiting and cost control for tool calls
• Practice: Build agent that orchestrates 5+ tools for a complex task

Week 4: Planning, Reasoning, and Multi-Agent Systems¶

Day 1-2: Advanced Reasoning¶

• Study chain-of-thought prompting for agents
• Learn tree of thought for multi-path reasoning
• Implement self-consistency for reliable decisions
• Practice inner monologue patterns
• Reference: NVIDIA Build - Test reasoning models

Day 3-4: Task Decomposition and Planning¶

• Study hierarchical task decomposition
• Implement dynamic re-planning based on results
• Learn convergence criteria (when to stop iterating)
• Practice planning for multi-step workflows

Day 5-7: Multi-Agent Systems¶

• Design multi-agent architectures (roles, communication)
• Implement agent-to-agent message passing
• Study orchestration patterns (sequential, parallel, hierarchical)
• Build a multi-agent debate system
• Lab: Build a 3-agent system with specialized roles

Week 5: Safety and Guardrails Deep Dive¶

Day 1-2: NeMo Guardrails Advanced¶

• Master Colang 2 syntax and features
• Implement complex dialog rails
• Build retrieval rails for RAG-based agents
• Create custom action handlers
• Reference: Colang 2 Overview

Day 3-4: Agent Safety Patterns¶

• Implement tool call validation and sandboxing
• Build human-in-the-loop approval workflows
• Study jailbreak detection and prevention
• Practice cascading failure prevention (timeouts, circuit breakers)

Day 5-7: Safety Testing¶

• Red-team your agents with adversarial prompts
• Test guardrails against known attack patterns
• Validate safety across diverse input scenarios
• Document and address discovered vulnerabilities
• Lab: Comprehensive safety audit of an agent system

Phase 3: Production and Exam Prep (Weeks 6-8)¶

Week 6: Production Deployment¶

Day 1-2: Scaling Agent Systems¶

• Deploy agent backend on Kubernetes with auto-scaling
• Configure load balancing for multiple NIM instances
• Implement queue-based processing for tool calls
• Study capacity planning for agent workloads

Day 3-4: Monitoring and Observability¶

• Set up monitoring for agent success/failure rates
• Track tool call patterns and token usage
• Implement distributed tracing for agent workflows
• Build alerting for unusual agent behavior

Day 5-7: Testing and Operations¶

• Write unit tests for tools and integration tests for workflows
• Implement load testing for agent infrastructure
• Practice blue-green deployments for model updates
• Build incident response procedures
• Lab: Deploy and monitor a production-ready agent system

Week 7: Review and Practice¶

Day 1-3: Domain Review¶

• Review all five domains systematically
• Create flashcards for key patterns and concepts
• Re-read NIM and Guardrails documentation for weak areas
• Complete any unfinished labs

Day 4-5: Practice Questions¶

• Work through scenario-based practice questions
• Focus on multi-domain questions
• Review answers and identify remaining gaps

Day 6-7: Gap Analysis¶

• Deep-dive into weakest areas
• Re-read documentation for problem areas
• Practice explaining agent patterns out loud

Week 8: Final Preparation¶

Day 1-3: Intensive Review¶

• Review fact sheet and architecture patterns
• Practice rapid concept identification
• Focus on safety and guardrails patterns

Day 4-5: Final Practice¶

• Full-length timed practice session (120 minutes)
• Review all incorrect answers
• Final review of weak areas

Day 6: Rest and Light Review¶

• Light review of fact sheet only
• Prepare exam logistics

Day 7: Exam Day¶

• Brief concept review (30 minutes max)
• Take the exam with confidence

Progress Tracking¶

Weekly Milestones¶

• Week 1-2: Understand agent patterns, set up NIM + Guardrails
• Week 3: Master tool use and function calling
• Week 4: Advanced reasoning and multi-agent systems
• Week 5: Safety and guardrails deep dive
• Week 6: Production deployment and operations
• Week 7: Practice questions and gap analysis
• Week 8: Final review and exam

Self-Assessment Questions¶

• Can I explain ReAct, Plan-and-Execute, and Reflexion patterns?
• Can I design a function calling schema for a complex tool?
• Do I know how to implement NeMo Guardrails with Colang?
• Can I architect a multi-agent system with proper orchestration?
• Do I understand production scaling and monitoring for agents?