Snowflake is accelerating the path from raw data to predictive insights with its new agentic ML capabilities, powered by Cortex Code. This AI coding agent, now generally available, aims to automate the traditionally slow and manual machine learning pipeline.
Data science teams can now use natural language prompts to develop production-ready ML solutions directly on Snowflake's platform, where their governed data resides. This eliminates context switching and leverages platform awareness for increased efficiency. First National Bank of Omaha, for instance, has seen a 10x productivity boost in forecasting and anomaly detection tasks.
The agentic approach democratizes ML, allowing non-technical teams to explore concepts and collaborate more effectively. Kargo, a creative performance platform, uses Cortex Code to enable its marketplace strategy team to test and scope new ideas, fostering deeper integration with data science efforts.
Cortex Code streamlines the entire ML lifecycle, from design and implementation to optimization. It intelligently triggers specialized skills for tasks like model training, inference deployment, distributed training, hyperparameter tuning, and performance monitoring.
This automation frees up data scientists to focus on higher-impact initiatives, leveraging their domain expertise rather than getting bogged down in documentation or debugging. Tasks like feature engineering, which were previously time-consuming, can now be rapidly iterated upon with simple prompts.
Snowflake claims significant performance advantages, with training up to three to seven times faster than open-source libraries and inference latency for XGBoost models up to 10x faster than legacy cloud providers. This is achieved through seamless scaling across CPUs and GPUs with built-in optimizations on the Snowflake ML platform.
Getting started is designed for flexibility, with Cortex Code accessible via the Snowsight UI or through a CLI for integration with tools like VS Code. Snowflake Notebooks provide a Jupyter-based environment for these workflows.
The platform aims to simplify the complex process of getting ML models into production, enabling faster iteration and quicker impact realization, helping teams to automate predictive insights.
