NVIDIA Certified Associate Generative AI LLMs (NCA-GENL) Cert Prep
9h 23mBeginner2026-06-04
Authors
Packt Publishing
Course details
Generative AI and large language models (LLMs) are reshaping how organizations build intelligent applications, and NVIDIA's tools are at the center of that shift. In this course, explore the foundations of AI and machine learning, transformer architecture, model selection, and customization techniques using NVIDIA's AI stack, including TensorRT, RAPIDS, and NeMo. Learn how to train and customize models for real-world use, discover ethical AI practices, and gain hands-on experience with deep learning workflows. By the end of this course, you’ll be prepared to sit for the NCA-GENL certification exam and ready to apply your skills to practical AI projects.
Concepts
Welcome
- Introduction
Why Generative AI Matters for Companies and You
- Why companies invest in generative AI and why it matters to you
NVIDIA Certification Details
- Know about certification tracks
- NCA-GENL certification
Module 1 - AI Infrastructure
- What we will cover
- AI and ML stack
- AI infrastructure
- GPUs for AI and ML
- GPU vs. CPU
- GPU architecture
- Memory pooling
- Cloud vs. on-premises
- Network infrastructure
- Storage infrastructure
- Module 1 - Summary
Module 2 - AI and ML Fundamentals
- AI and ML
- Deep learning
- A simple use case
- What is a model
- Training a model
- Module 2 - Summary
Module 3 - Generative AI & LLM
- What is generative AI
- Foundation model
- Large language model
- Road to generative AI
- Transformers
- Module 3 - Summary
Module 4 - Transformer Architecture
- Transformers - The engine behind modern AI
- How transformers work
- Tokenization
- Encoding
- Word embedding
- Decoding
- Output
- Positional encoding
- Transformer explainer
- Attention mechanism
- Multi-head attention
- Encoder decoder
- Which one to use
- Prediction strategy
- Module 4 - Summary
Module 5 - Model Selection
- How to pick the right AI foundation model
- Model selection process
- SLM vs. LLM
- Model evaluation
- Metrics that matters
- Metrics comparison
- Cross validation
- A B testing
- Selecting best model
- Module 5 - Summary
Module 6 - Model Customization
- Why customize a model
- Prompt engineering
- System prompt
- Retrieval-augmented generation (RAG)
- Chunking
- Transfer learning
- Transfer Learning Approaches
- Transfer learning use cases
- Fine-tuning
- Knowledge distillation
- Accuracy on a validation set
- Module 6 - Summary
Module 7 - Model Training
- Model training analogy
- Beer or wine
- Order of steps
- Data collection
- Data processing - RegEx
- Data processing - EDA
- EDA Techniques
- Model training
- Model training phases
- Model Evaluation
- Model deployment
- Options for deployment
- ONNX format
- Quantization
- Post deployment
- Module 7 - Summary
Module 8 - NVIDIA Ecosystem
- NVIDIA ecosystem
- NVIDIA GPUs
- CUDA
- NVIDIA SMI
- NVIDIA RAPIDS
- How RAPIDS work
- NVIDIA TensorRT
- TensorRT vs. quantization vs. ONNX
- NVIDIA NeMo
- NVIDIA Triton
- Dynamic batching
- NGC catalog
- Other integration
- Module 8 - Summary
Module 9 - Ethical AI
- Ethical AI
- Building trustworthy AI
- Confidential computing
- NVIDIA NeMo Guardrails
- Embedding classifier
- Data lineage tracking
- Red teaming
- Data augmentation
- Accountability
- Explainable AI
- Module 9 - Summary
Module 10 - Additional Topics
- Forward and back propagation
- Vanishing gradients
- Challenges of vanishing gradients
- Neuron
- Inside a neuron
- Forward Diffusion
- Reverse diffusion
- Rapid application development (RAD)
- Hugging Face transformers
- Module 10 - Summary
- Is it a cat
- ReLU vs. sigmoid
- Layer normalization
- Named-entity relationship
- Diffusion algorithms
- Forward and reverse diffusion
Exam Tips
- Prepare for the certification exam