NCA-GENM learning guides

One short guide per official objective, grouped by exam domain. Each point is backed by a quote from an NVIDIA source. Terms link to the glossary the first time they appear.

Experimentation

25% of the exam · 5 objectives

1.1 Developing and testing multimodal models
1.2 Managing and preprocessing multimodal data
1.3 Explainability with multimodal models
1.4 Testing multimodal data quality
1.5 Testing models for accuracy

Core Machine Learning and AI Knowledge

20% of the exam · 10 objectives

2.1 Training stability
2.2 Multimodal loss functions
2.3 Machine learning fundamentals
2.4 Nonsequential networks and residual connections
2.5 Statistics for evaluating pipelines
2.6 Multimodal transfer learning
2.7 Emerging multimodal trends
2.8 Energy-efficient, trustworthy models
2.9 Prompt engineering principles
2.10 TensorFlow and PyTorch

Multimodal Data

15% of the exam · 7 objectives

3.1 Scalability, performance and reliability
3.2 RAG, chatbots and summarizers
3.3 Python natural-language packages
3.4 Identifying data, hardware and software components
3.5 Monitoring data collection and experiments
3.6 Python packages for traditional ML
3.7 Writing software components and scripts

Software Development

15% of the exam · 5 objectives

4.1 Working with clients on requirements
4.2 Best practices and software quality
4.3 Prompting generative models
4.4 U-Nets: from noise, and as an autoencoder
4.5 CLIP for text-to-image

Data Analysis and Visualization

10% of the exam · 4 objectives

5.1 Insights from large datasets
5.2 Attention maps
5.3 Graphs and charts
5.4 Relationships, trends and confounders

Performance Optimization

10% of the exam · 4 objectives

6.1 Efficiency versus accuracy
6.2 Hyperparameter tuning
6.3 Transfer learning for performance
6.4 Training optimization

Trustworthy AI

5% of the exam · 4 objectives

7.1 Ethical principles of trustworthy AI
7.2 Privacy and consent
7.3 NVIDIA technology for trust
7.4 Minimizing bias