JFrog integrates with Hugging Face, Nvidia; intros JFrog ML
The partnership aims to give developers a clearer guide on safe models. It also aims to ensure enterprise-grade safety and security with the deployment of open models.
MLOps and supply chain software vendor JFrog, building on past partnerships with Hugging Face and Nvidia, on Tuesday introduced a new supply chain platform for AI and traditional software development.
The vendor is integrating with AI and data science platform Hugging Face to provide high trust to data scientists, machine learning (ML) engineers and developers by displaying a "JFrog Certified" checkmark that clarifies which model is safe to use. With the partnership, JFrog Advanced Security and JFrog Xray will scan AI and ML model artifacts in the Hugging Face Hub for threats within each model.
With AI chip giant Nvidia, JFrog revealed that its integration with Nvidia NIM microservices is now generally available. The integration lets enterprises deploy and manage foundation models such as those from Meta and Mistral while maintaining enterprise-grade security and governance controls in their software supply chain.
JFrog also unveiled JFrog ML, an MLOps capability. JFrog ML was initially released under the Qwak MLOps platform. JFrog acquired Qwak in June, and while the capability was under Qwak, it was not incorporated with the JFrog platform. Now, the capability is completely incorporated in the JFrog platform. It provides organizations with a structured framework that supports them and helps move models from the experimental stage to implementation, JFrog said.
Responding to customers
While the integrations with Hugging Face and Nvidia as well as the introduction of JFrog ML address some of the trust and security challenges enterprises face when deploying AI models, they are also a way for the MLOps provider to respond to its customers.
It is good to see them deepen partnerships and integrations to bring robust solutions to their customers and the open source community.
Katie NortonAnalyst, IDC
JFrog initially revealed the integration with Nvidia last year. The integration enables users to host the NIM models with the JFrog Platform. It also reduces the challenges that come with deploying NIM models and uses GPU-optimized configuration to speed up the model performance, IDC analyst Katie Norton said.
The Hugging Face integration addresses the reality of open source as a point of attack for malicious attackers. It also helps provide the open source community a service that increases visibility and makes the choice of secure models easier for developers, she added.
"It is good to see them deepen partnerships and integrations to bring robust solutions to their customers and the open source community," Norton said.
However, she questioned whether JFrog is willing to be held liable if a model is marked as "JFrog Certified" but later found to have a vulnerability or malicious aspect that JFrog did not uncover.
Securing MLOps and AI systems
Despite this possibility, JFrog has acted as an early mover in securing MLOps and AI models, which is essential, Norton said.
"Like any other software component, ML models must be stored, built, traced, versioned, signed, secured and delivered across systems to achieve AI at scale," she said. "Implementing DevOps best practices within a unified solution that includes MLOps can help organizations meet these market expectations."
For JFrog, the goal is to focus on enterprise IT security.
"Our goal is to give enterprises the trust and the ability to integrate faster," said Yuval Fernbach, vice president and CTO of MLOps at JFrog. "The same story that JFrog was founded for years ago, about the ability for companies to develop faster and be more agile and be more secured, we're now doing that not just for software, but ML."
Esther Shittu is an Informa TechTarget news writer and podcast host covering artificial intelligence software and systems.
