NVIDIA Certified Professional: Agentic AI NCP-AAI Prep Course
A structured course covering every NCP-AAI exam domain, delivered as 58 lessons across 10 modules. Work through it in order, or jump to the module for the domain you are weakest in.
Threads through the course
Six throughlines that run across modules, so the course reads as continuous rather than as fourteen separate topics.
- The measurement thread
- The weights thread
- The infrastructure thread
- The efficiency thread
- The control thread
- The core-concepts thread
- The generative pipeline thread
- The multimodal-measurement thread
- The compute-efficiency thread
- The trust and safety thread
- The resilience thread
- The memory and grounding thread
- The NVIDIA stack thread
- The oversight thread
- The model-efficiency thread
- The adaptation-strategy thread
- The regression-measurement thread
- The guardrail-limits thread
Course outline
58 lessons delivered as 58 sessions · 36.1h from 36.1h designed · 58 written
Module 2 · 6 lessons · 4.9h
Agent Development
Given a model that can call tools, how do you keep the thing alive when a tool call, a network, or a whole dependency fails?
obj 2.1 · 2.2 · 2.3 · 2.4 · 2.5 · 2.6The resilience thread
Why this weighting
Tied with Agent Architecture for the heaviest weight (15%). The Retry and Circuit Breaker state machines are the most testable, most confusable pieces of this domain.