Unmatched Collaboration for Data & AI Products: What’s New

Posted on by

Getting different teams, business units and even companies to work together toward a common goal not only maximizes efficiency, but drives innovation. Effective collaboration on data and AI has never been more closely tied to success. 

At Snowflake, we’re removing the barriers that prevent productive cooperation while building the connections to make working together easier than ever. Whether it’s helping you share data and AI models internally or build and distribute groundbreaking AI and advanced solutions, Snowflake is committed to helping you weave collaboration into the fabric of your business. Here’s a closer look at some of the innovations we’ve introduced to make all of this even easier while keeping privacy and governance in mind.

Cross-cloud collaboration and discovery

Strengthening our established leadership in cross-cloud collaboration capabilities, Snowflake is introducing more ways to securely share data, apps and AI models both internally and externally. 

Easily share apps, data and AI models across your enterprise

Where Snowflake Marketplace has helped companies discover new high-quality data, apps and AI products, Internal Marketplace helps maximize their use. Internal Marketplace (now generally available) provides an all-encompassing directory of data products available within an organization. With everything available for discovery on a single pane of glass, it’s easy for data consumers to find and access the data, AI models and apps they need, when they need them. Plus, you don’t have to create (or pay for) multiple copies of the same resource for different departments, which streamlines management and provides a better security posture.

Evaluating these products now is incredibly easy, too, with Copilot for Listings (private preview). This feature allows users to ask questions in natural language to find out whether the attached contents of a product listing are relevant and valuable to their work.

Share fine-tuned LLMs

Snowflake now supports seamless sharing of fine-tuned LLMs and exploring of model weights with Secure Model Sharing (public preview). This makes it easier to collaborate on generative AI use cases and deploy models into production, as the recipient can see the fine-tuned model in their Snowflake Model Registry UI. Moreover, since no actual data is copied or transferred between accounts — only Snowflake’s services layer and metadata store are used — sharing models reduces the risk of data exposure.

Build more sophisticated data and AI products while remaining cost-effective

Innovative companies are bringing their latest technologies to Snowflake, offering cutting-edge products that don’t require moving data or carrying out complex orchestrations. Companies can share and monetize advanced AI and ML models and applications through Snowflake Native Apps, now fully integrated with Snowpark Container Services to support more use cases. And with the advanced cost-optimization tools available as part of the Snowflake secure sharing solution, customers can share data and collaborate across clouds without worrying about cross-cloud egress costs.

Snowflake Native App Framework integrates with Snowpark Container Services

Since its introduction in 2022, the Snowflake Native App Framework has provided a secure and cost-effective way to build, distribute and monetize data and AI products that seamlessly integrate with Snowflake, simplifying app development for hundreds of companies. Now, with Snowpark Container Services generally available and fully integrated into the Snowflake Native App Framework, you can create even more sophisticated apps and deploy them to customers on AWS (generally available) and Azure (public preview). 

Snowpark Container Services (now generally available on Azure as well as AWS) gives developers the flexibility to easily and securely deploy complex functionality — from custom frontends and large-scale ML training and inference to open source and homegrown models — all securely within Snowflake (learn more in this blog post). By packaging and registering a container image and then creating a Snowflake Native App package with those container details, developers can:

  • Build robust products in any programming language

  • Flexibly use GPUs or high-memory CPUs, depending on their needs

  • Distribute apps seamlessly across clouds and regions

  • Remove the need to move data outside your security perimeter

Providers are building sophisticated apps in all sorts of interesting categories, so it’s highly likely that if you’re looking for a certain functionality, there’s an app for that in Snowflake Marketplace. For example, you can:

  • Perform complex data transformation with dbt

  • Add autonomous AI agents to your workforce with Genesis

  • Incorporate advanced graph analytics from RelationalAI 

  • Leverage predictive AI from Kumo AI

  • Turn images and videos into visual intelligence with LandingAI

Build and monetize AI products

With the Snowflake Native App Framework, registered AI/ML models can now be packaged up, shared, and monetized. Using the Snowpark ML Modeling API, providers can create, train and log models in the Snowflake Model Registry, then export them into a Snowflake Native App package for listing. 

This unlocks new revenue streams, allowing organizations to easily and securely capitalize on their model-development efforts and provide high-value AI and ML products on Snowflake Marketplace for others to use or buy. With more product options to choose from, consumers have a greater chance of finding the right solution for their needs. 

Removing duplicative egress fees 

To make cross-cloud collaboration more cost-effective, Snowflake is introducing the Egress Cost Optimizer (generally available soon), which essentially removes duplicative egress costs when sharing data or apps to multiple cloud regions. It helps providers gain control over data product delivery costs and unlock demand from other cloud regions.

With Egress Cost Optimizer, customers will pay egress fees only the first time data is shared to an account in a different cloud region. There are no incremental egress charges to share that same data to other accounts in that region or any other cloud regions, and no egress costs when using Snowflake Secure Data Sharing within the same cloud region.

Snowflake’s patented cross-cloud technology uses a replication-based approach to enable access to data in remote regions. That means Snowflake manages the replication process using Cross-Cloud Auto-Fulfillment, as opposed to a remote query-based approach, where the provider is charged small egress fees for each and every query. The latter can quickly become costly, and for large data sets, query performance is often slowed by the constant data movement. With a replication-based approach and Egress Cost Optimizer, providers can achieve high query performance on the shared data while also gaining control over delivery costs.  

Snowflake’s commitment to security lets you build and use apps with confidence  

Sharing and collaboration can come with some inherent governance challenges, especially when data or apps are exposed to the public internet without security mechanisms in place. That’s why Snowflake works to provide both providers and consumers with tools to help ensure their data, applications and models remain safe and secure. 

Snowflake Native App support for secure deployments 

The Snowflake Native App Framework now supports Virtual Private Snowflake (VPS) and AWS PrivateLink, enabling enhanced security and privacy for app providers. This feature (generally available) will help organizations build trust with high-security clients and expand their offerings into regulated markets. As it stands, customers who require robust levels of security now have access to more than 220 Snowflake Native Apps on Snowflake Marketplace to help accelerate their data and AI initiatives.

Compliance badges

Snowflake Native App providers will also be able to display third-party compliance certifications, such as SOC 2, as part of their app listing (private preview). Once Snowflake has successfully reviewed their compliance documents, providers can update their listing to include relevant badges, enabling app consumers to quickly and confidently see whether a solution is capable of meeting their compliance requirements at a glance.

Learn more

Snowflake’s unified platform simplifies the inherent complexities of collaboration through privacy-preserving data sharing capabilities and by allowing you to bring applications to the data rather than the other way around. Not only is it more efficient and more secure, it’s also simply easier to collaborate within an organization and with external partners with Snowflake. 

For more about Snowflake collaboration capabilities and Snowflake Native Apps, watch our BUILD 2024 What’s New session, check out our developer resources and the Snowflake Native App Bootcamp, and register for our Snowflake Marketplace virtual event to learn how to accelerate AI initiatives with data, apps and AI products from third-party providers.