Christian Kleinerman, Snowflake & Manuvir Das, NVIDIA | Snowflake Summit 2023
summary
Announces a Snowflake integration that runs NVIDIA’s stack inside Snowflake containers enabling model training, fine tuning, prompt tuning, retrieval augmentation and reinforcement learning directly on enterprise data. Uses containerized NeMo stack and GPU acceleration with an option for serverless inferencing via the Predict V2 API, plus a security model that keeps customer training data and resulting models isolated. Demonstrates recommender and anomaly detection pipelines built from Snowflake-curated datasets using Morpheus and GPUs for faster, cheaper runs.
AI-written summary · about
chapters (11)
- 0:00Introduction and partnership
- 1:11Why the NVIDIA-Snowflake partnership
- 4:03Containerized NVIDIA stack in Snowflake
- 6:11Customizing AI models with data
- 9:48Integrating LLMs with Snowflake workflows
- 12:05Data pipelines for AI training
- 13:28Democratizing AI applications
- 15:35Snowflake leapfrogging with gen AI
- 16:58Turbocharging Snowflake with GPUs
- 19:04Admin experience with NVIDIA stack
- 20:09Closing and next sessions
