Just registered for NVIDIA NCP-AAI certification?

Here's your complete study guide.

**Exam details:**
• 60-70 multiple choice questions
• 120 minutes (2 hours)
• Pass score: 70% (need 42-49 correct)
• Cost: $200
• Valid for 2 years

**What's tested (10 domains):**

**HIGH-WEIGHT (43%):**
• Agent Architecture and Design - 15%
• Agent Development - 15%
• Evaluation and Tuning - 13%

**MEDIUM-WEIGHT (27%):**
• Cognition, Planning, and Memory - 10%
• Knowledge Integration and Data - 10%
• NVIDIA Platform Implementation - 7%

**LOW-WEIGHT (30%):**
• Run, Monitor, and Maintain - 7%
• Deployment and Scaling - 5%
• Safety, Ethics, and Compliance - 5%
• Human-AI Interaction and Oversight - 5%


📊 6-WEEK STUDY PLAN:

**Weeks 1-2:** Architecture + Development (30% of exam)
- ReAct frameworks, LangGraph workflows
- Tool integration, error handling
- BUILD: 3 agents with different patterns

**Week 3:** Evaluation + Cognition (23% of exam)
- Evaluation pipelines, A/B testing
- CoT reasoning, memory management
- EVALUATE: Your agents

**Week 4:** Data + NVIDIA Platform (17% of exam)
- RAG systems, vector databases
- NeMo Guardrails, NIM, TensorRT-LLM, Triton
- BUILD: RAG system with guardrails

**Week 5:** Deployment + Operations (12% of exam)
- Containerization, monitoring, logging
- DEPLOY: One agent to production

**Week 6:** Safety + Human Interaction + Review (10%)
- Compliance, bias mitigation
- REVIEW: All topics

---

📚 STUDY RESOURCES:

**NVIDIA Official Courses (FREE - MUST TAKE):**
→ Building Agentic AI Applications with LLMs
→ Building RAG Agents with LLMs
→ Deploying RAG Pipelines for Production at Scale
→ Evaluating RAG and Semantic Search Systems

**Documentation:**
→ LangGraph (complete docs)
→ NVIDIA NeMo
→ LangChain agents
→ Agent Intelligence Toolkit

**Papers:**
→ ReAct paper
→ "Understanding the Planning of LLM Agents: A Survey"

**Hands-On (MOST IMPORTANT):**
→ Build 5+ agents
→ Implement RAG with evaluation
→ Deploy and monitor

---

🎯 EXAM STRATEGY:

**Focus on high-weight topics:**
Architecture (15%) + Development (15%) + Evaluation (13%) = 43% of exam

**NVIDIA platform is critical:**
NeMo Guardrails, NIM, TensorRT-LLM, Triton = 7% of exam

**The exam tests PRODUCTION knowledge:**
Not "What is X?" but "How would you solve Y in production?"

**Timeline:**
4-6 weeks if working full-time with agents
2-3 weeks if building production agents daily
8-10 weeks if newer to the field