Kumo Brings Deep Learning to Snowflake Data Cloud via Snowpark

Kumo, a deep learning platform for relational data, today announced at Snowflake Summit 2023 that it will integrate deep learning capabilities directly into the Snowflake Data Cloud through Snowpark Container Services.

Snowflake recently introduced Snowpark Container Services, which extend the capabilities of Snowpark. This update enables organizations to run third-party software and full-stack applications within Snowflake accounts.

According to Snowflake, this integration will allow customers to maintain data security and eliminate the need for data movement while using cutting-edge tools to maximize their data’s potential.

Additionally, Snowpark Container Services includes GPU support so data science and machine learning teams can accelerate development and close the gap between model deployment and consistent data security and governance across the AI/ML lifecycle. can be filled.

Kumo was one of the early adopters of Snowpark Container Services, using the technology to deploy advanced neural networks for the enterprise.

Kumo’s predictive AI platform, powered by graph neural network (GNN) technology, enables developers, data scientists, analysts, and business owners to create and implement highly accurate predictions in production.

Graph Neural Networks and AI

Traditional machine learning requires extracting data from a data warehouse or lake and then manually developing and tuning features. The new integration is now available in private his preview, allowing co-users to work directly with live her Snowflake tables. Generate predictions. and saves the results as an additional table in her Snowflake.

“The new integration uses graph neural network technology to enable Kumo’s AI services on relational tables in the cloud without the intermediate steps found in traditional machine learning, such as training set generation and feature engineering. directly,” the CEO of Kumo told VentureBeat.

Josifovski emphasized that users can create and run queries that provide forecasts, reflecting the process of querying historical data for analysis, without having to export data from the Snowflake environment.

The announcement follows a recent partnership between Nvidia and Snowflake that allows customers to customize generative AI models through the cloud to suit their specific enterprise requirements.

This integration allows organizations to develop generative AI applications using their own data within Snowflake’s Data Cloud environment, eliminating the need to transfer data externally.

Facilitate deep learning-based predictive analytics on the cloud

According to Kumo’s Josifovski, Snowpark Container Services will enable customers to leverage Kumo’s predictive AI services directly within Snowflake to perform graph learning predictions on enterprise data.

“A long-standing question about machine learning and data warehousing has been where ML processing is performed. We will extend the use of machine learning and prediction to everyone with access to the cloud,” Josifovski told VentureBeat. “This is done in a single security program, which makes it much simpler than doing it in multiple security programs.”

Modern AI techniques rely heavily on linear algebra computations, which are highly compatible with GPU processing. Previously, taking advantage of GPUs required Kumo to extract data from customer accounts and process it externally. With this integration, all data processing, including GPU processing, takes place directly within the customer’s Snowflake account.

“An approach that does not require training sets or feature engineering significantly shortens the AI/ML lifecycle,” he added. “We free data scientists from repetitive and tedious tasks so they can focus on the higher-level tasks of defining good prediction tasks, evaluating outcomes, and finding the best way to derive business value from predictions. I aim to do so.”

Through this product, the company introduced a unique capability: a deep learning-driven relational data GNN.

These deep learning-driven GNNs can learn from graphs and associated attributes determined by non-key columns of data. Once the graph is built, multiple AI/ML tasks can be efficiently trained on the same graph without creating separate training sets or numerous engineering functions.

Kumo also provides scalable and innovative autoML algorithms that alleviate the tedious process of hyperparameter tuning.

“GNNs are very effective for a variety of prediction problems, but they are also difficult to implement, scale, and make efficient. Using Kumo’s AI platform, you need to be familiar with GNNs and creating optimization tasks for graph Kumo has implemented a predictive query language to specify AI/ML tasks,” said Josifovski.

Streamlining Predictive Analytics for Citizen Developers

Josifovski said predictive AI/ML currently requires highly skilled experts with narrow expertise. Since the lifecycle involves feature experimentation, it requires substantial infrastructure support for training and inference (scoring).

He explained that the purpose of the new integration is to give users a streamlined workflow regardless of their data science proficiency.

Predictive graph learning can then be easily applied to various business domains such as customer acquisition, loyalty, retention, personalization, and fraud detection. His company claims the entire AI-based analysis can be completed in hours.

“Kumo allows users to query relational data without requiring a deep understanding of AI/ML concepts, while giving trained data scientists control over training and inference,” says Josifovski. says Mr. “In this way, the platform can be used by a wider audience in the same way that data warehouses are used for analytics today.”

Additionally, Kumo emphasized that the native integration with Snowflake makes the product easy to install and use without requiring security or legal privacy reviews. This reduces barriers and significantly accelerates time to value.

The company believes this will encourage experimentation and adoption of detailed predictions, enabling and improving practices such as customer acquisition, personalization, entity resolution, and other predictive tasks.

“In the enterprise, many teams issue SQL queries on data warehouses to get analytics that experts use to plan future actions,” Josifovski told VentureBeat. “Kumo will enable users to obtain actionable predictions in an automated manner without the need for expert interpretation.”