NVIDIA Certified Associate: Generative AI Multimodal Certification Training Course

Core Machine Learning and AI Knowledge (20%)

Knowledge of algorithms, conventions, and techniques that allow computers to learn from and make predictions or decisions

• Control training stability in multimodal settings.
• Develop content introducing multimodal loss functions.
• Apply machine learning fundamentals (feature engineering, model comparison, cross-validation).
• Explain nonsequential neural networks and residual connections.
• Design statistical analyses for evaluating multimodal pipelines.
• Develop content for multimodal-specific transfer learning.
• Track emerging multimodal trends and technologies.
• Contribute to the design, development, and deployment of energy-efficient, trustworthy multimodal AI models.
• Apply prompt engineering principles to craft prompts that achieve desired results.
• Work with deep learning frameworks such as TensorFlow or PyTorch.

Data Analysis (10%)

Inspecting, cleansing, transforming, and modeling data with the goal of discovering useful information, informing conclusions, and supporting decision-making.

• Awareness of the process of extracting insights from large datasets using data mining, data visualization, and similar techniques.
• Develop content for attention maps in multimodal settings.
• Create graphs, charts, or other visualizations to convey the results of data analysis using specialized software.
• Identify relationships and trends or any factors that could affect the results of research.

Experimentation (25%)

The study of how to perform, evaluate, and interpret experiments, including AI model evaluation and the evaluation of various model architectures.

• Assist in developing and testing multimodal AI models.
• Manage and preprocess data from various sources.
• Use multimodal models to improve explainability.
• Test data quality and consistency in a multimodal setting.
• Test AI models to ensure their accuracy and effectiveness.

Multimodal Data (15%)

Multimodal data involves the integration, curation, and quality assessment of diverse data types such as text, images, audio, time-series, and geospatial information, while also addressing challenges related to missing or incomplete information across these different modalities.

• Assist in the deployment and evaluations of model scalability, performance, and reliability under the supervision of a senior team member.
• Build LLM use cases such as retrieval-augmented generation (RAG), chatbots, and summarizers.
• Familiarity with the capabilities of Python natural language packages (spaCy, NumPy, vector databases, etc.).
• Identify system data, hardware, or software components required to meet user needs.
• Monitor the functioning of data collection, experiments, and other software processes.
• Use Python packages (spaCy, NumPy, Keras, etc.) to implement specific traditional machine learning analyses.
• Write software components or scripts under the supervision of a senior team member.

Performance Optimization (10%)

Performance optimization in AI entails refining multimodal AI models for energy efficiency, trustworthiness, and accuracy through design contributions, transfer learning content development, supervised training enhancements, hyperparameter tuning, rigorous testing, and computational advancements.

• Enhance computational efficiency and improve the accuracy of outputs in AI models.
• Optimize the performance of AI models, including tuning hyperparameters.
• Develop content for multimodal-specific transfer learning.
• Assist in model training and training optimization under the supervision of a senior team member.

Software Development (15%)

Design and implement neural network architectures, such as U-Nets for generative image tasks, integrate text-to-image AI models like CLIP, and apply prompt engineering to refine and direct the generative capabilities of these systems. Includes familiarity with NVIDIA SDKs such as NVIDIA NeMo™ and NVIDIA Triton™ Inference Server.

• Collaborate with the client during requirements acquisition, data gathering, progress reporting, deployment, and integration.
• Ensure adherence to best practices and maintain high standards of software quality and reliability.
• Use prompt engineering to better influence the output of generative AI models.
• Build a U-Net to generate images from pure noise and as a type of autoencoder.
• Generate images from English text prompts using CLIP, and use CLIP to train a text-to-image diffusion model.

Trustworthy AI (5%)

Create and assess ethical, energy-conscious, and reliable artificial intelligence systems that are capable of interpreting and integrating various forms of data, ensuring that they’re designed and applied in a manner that’s transparent, fair, and verifiable.

• Describe the ethical principles of trustworthy AI.
• Describe the balance between data privacy and the importance of data consent.
• Describe how to use NVIDIA and other technologies to improve AI trustworthiness.
• Describe how to minimize bias in AI systems.

Exam details

• Level: Associate
• Questions: 50-60 multiple choice
• Length: 1 hour
• Price: $125
• Language: English
• Scoring: pass/fail, with no score reported
• Validity: 2 years from issuance