Enterprise machine learning faces significant challenges in security, reproducibility, and portability between environments. Jozu is addressing these core issues by pioneering two key open-source projects: the ModelPack specification and KitOps. These initiatives, backed by the Cloud Native Computing Foundation (CNCF), represent the first effort to bring the security and portability standards of containers to AI/ML workflows, making model delivery as reproducible and verifiable as any container deployment.
Jozu is a leading developer of the ModelPack specification and the open-source KitOps project.
ModelPack extends the OCI standard to support large AI artifacts like models and datasets.
KitOps packages these artifacts into versioned, portable "ModelKits."
The approach makes AI deployments cryptographically verifiable and reproducible.
KitOps has seen over 140,000 downloads in the past year.
Jozu now offers enterprise support and management through its Jozu Hub platform.
AI and ML teams have long struggled with the logistical challenge of moving complex projects—comprising models, datasets, and code—securely from development to production. The ModelPack specification directly addresses this by extending the widely adopted Open Container Initiative (OCI) standard to natively support the large file types common in AI. This creates a universal, open standard for packaging and versioning AI projects.
KitOps is the reference implementation of the ModelPack specification. It assembles all components of an AI project into a single, versioned unit called a ModelKit. These ModelKits integrate seamlessly with existing container registries and Kubernetes infrastructure, requiring no new tooling. This allows teams to move models through development, testing, and production with the same confidence and ease they have with containers, ensuring deployments are cryptographically verifiable and reproducible.
Early enterprise adopters are already seeing significant results from standardizing on this new approach. Global logistics leader DSV uses KitOps to accelerate AI delivery across its 90+ offices. A security-focused German systems integrator reported a 41% acceleration in its delivery cycles while simultaneously automating adherence to its stringent security and compliance requirements, demonstrating the dual benefit of speed and security.
The project has garnered significant industry backing from major players like PayPal, ByteDance, ANT Group, and Red Hat. As Görkem Ercan, CTO and Co-Founder of Jozu, stated, “ModelKits make AI deployments cryptographically verifiable, compliant, and reproducible across any Kubernetes environment.” This initiative is positioned to provide the AI/ML ecosystem with the same foundational standards that containers provided for application development.
Jozu provides production operations infrastructure for enterprise AI through secure model packaging and deployment. Built on open source KitOps (CNCF Sandbox) and the ModelPack specification, Jozu's Kubernetes-native platform integrates with existing container registries and delivers tamper-proof deployments with complete audit trails. Trusted by government agencies and global enterprises for both connected and air-gapped environments.
TechIntelPro empowers B2B technology decision-makers with the latest industry trends, deep insights, leadership perspectives and actionable intelligence to drive both organizational and professional milestones—and thrive on the tech pulse.