The Neuron: AI Explained NVIDIA’s Kari Briski on How to Use NVIDIA Nemotron Open-Source AI
35 snips
Oct 15, 2025 Kari Briski, Vice President of Generative AI for Enterprise at NVIDIA, discusses the powerful capabilities of Nemotron, NVIDIA's open-source AI models. She explains the key differences between the Nano, Super, and Ultra tiers, emphasizing the importance of hardware specs. Briski dives into why businesses might prefer local deployment over cloud solutions for privacy and specialization. She also touches on the potential for Nemotron to enhance specialized applications in sectors like healthcare and banking, making AI more accessible and tailored for various industries.
AI Snips
Chapters
Transcript
Episode notes
Customize Models With Recipes And Distillation
- Use Nemotron recipes to distill, quantize, or alter architectures for your domain.
- Modify model internals and retrain to create smaller, faster models tailored to your needs.
Train With Synthetic Data And Gyms
- Use released datasets and synthetic data generation to save compute and widen model exposure.
- Build gym environments and RL pipelines to exercise models on diverse, task-specific scenarios.
Local Models Protect IP And Enable Specialization
- Running models locally preserves private IP, domain expertise, and data control.
- Enterprises need specialization beyond general cloud APIs to reach production-grade vertical applications.