Snowflake and NVIDIA announced their partnership on 26 June. The planned NeMo integration would fine-tune language models using customers’ Snowflake data, with chatbots and search among the proposed uses.
Archive photograph: © 2020 Will Buckner, via Wikimedia Commons. NVIDIA headquarters sign, 31 October 2020. CC BY 2.0. Cropped for display; original retained. Not a photograph of the partnership announcement. No endorsement implied.
Preview boundaries
The 27 June explanation puts Snowpark Container Services in private preview. Its recommender example combines RAPIDS/Merlin training, TensorRT optimisation and Triton inference.
Fine-tuning adapts existing models; inference runs them. Training a recommender doesn’t establish general availability for custom language models.
The design brings model development to data already in Snowflake accounts. Keeping work there doesn’t establish a privacy guarantee or replace access controls.
Reuters reports that financial terms weren’t disclosed. NVIDIA expects chip sales and paid AI Enterprise software to benefit.
Discover more from Sulat AI & Tech News
Subscribe to get the latest posts sent to your email.
