Extract Deeper Insights from Data with H2O.ai and Snowflake

Empowering Snowflake Users with AI Technology

Organizations are looking to uncover new meaningful insights from their data stored in the Snowflake Cloud Data platform through the use of AI and machine learning. However, many enterprises are challenged with the diverse skill sets and resources needed to take advantage of these technologies. The integration of H2O Driverless AI with Snowflake removes many of those barriers by using SQL commands in Snowflake for training and deploying predictive models in production. The H2O and Snowflake solution provides a cloud-agnostic platform so customers can work freely across any cloud for their AI initiatives.

Accelerate ROI on AI Initiatives

• Flexibility
  Cloud agnostic platforms avoid customer lock-in
• Performance
  10X performance improvement enables real-time insights
• Ease of Use
  Use SQL in Snowflake for predictive analytics
• Scalability
  Accelerate end-to-end model deployment
• Governance
  Model testing and validation using Snowflake Time Travel and Streams

Snowflake Integration Overview

Snowflake users who want to easily extract new and meaningful insights from all their data can apply automatic machine learning (AutoML) for business predictions. Providing those Snowflake users with H2O Driverless AI through familiar SQL commands within their existing workflows makes them both self-sufficient and more productive. This is made possible through the use of Snowflake External Functions – a feature available to all Snowflake customers. The end-to-end ML pipeline can now be executed by data analysts or data engineers using SQL statements within Snowflake so users don’t have to learn a new technology platform.

Training ML Models Using SQL in Snowflake

H2O enables Snowflake users to train or retrain ML models using simple SQL commands from within Snowflake. Once a model has been created it can be automatically deployed into production as part of the typical operations process. This removes the need to go to other systems or even teams to train or retrain models, saves the operations team valuable time, and keeps high performing models running in production.

Making Predictions in Snowflake Using SQL

With the Snowflake integration, predictive models act like they’re embedded in the SQL workflow, and execute from within the Snowflake process. There is no longer a need to extract the data from Snowflake to score data for making predictions, but rather the scoring of data to make predictions takes place straight from within Snowflake using SQL commands. DevOps no longer needs to setup and manage a batch process, which saves on setup time and management resources.