Vinay Sridhar

General-purpose LLMs are powerful but often expensive and not specialized for enterprise use. As AI scales, organizations need models that are cheaper, smaller and deeply aligned to proprietary data. In this session, we show you how to take advantage of comprehensive solutions for LLM customization, across all levels of complexity, directly in Snowflake. From APIs that simplify the workflow and provide easy experimentation loops to research-focused capabilities supporting reinforcement learning and advanced optimization, teams can start simple and evolve seamlessly, while ensuring they meet their quality, cost and latency goals.

Modern time series forecasting increasingly relies on transformer models, often deployed in scenarios where datasets are partitioned such as across product segments in retail, leading to the need to create a time-series model per segment, often 100s to 1000s at the same time. Training these...

Deploying machine learning models into production typically involves automated CI/CD pipelines. After experimentation, data scientists commit finalized code, initiating pipelines that replicate development steps on production data for consistency and scalability. This session introduces how Snowflake ML is designed to simplify ML workflow deployment by enabling remote execution directly from orchestration engines, including Snowflake Tasks. Come to this session to learn more about how you can productionize your end-to-end ML pipelines within Snowflake and configure them to trigger model retraining runs and visualize and monitor task execution graphs.