Jesse Williams · November 18, 2024

We’re submitting KitOps to the CNCF

Hey everyone, it’s been a busy few months for us and we’re overdue for a project update.

Last week we had a few of our team members attend KubeCon NA, where we had the opportunity to discuss the challenges we’re working on solving with KitOps.

AI and ML are becoming integral components of cloud-native application architecture, and the fragmented packaging and versioning of models and associated datasets, codebases, and parameters is causing delays, compliance risks, and project failures. Many of these fragmentation problems have already been solved for cloud native applications, thanks to standards like git and OCI that are used across all vendors’ tools and infrastructures.

AI models are different from today’s loosely coupled microservices. AI models have tight couplings with the datasets that trained and validated them, as well as their weights, parameters, inputs, and outputs. This creates complexity, which is only exacerbated by the lack of standardized packaging that forces bits and pieces to be imported, exported, and versioned separately across multiple repositories and proprietary vendor formats. In development these are frustrations, but as projects move to production they become material operational and business risks.

Having come from organizations like AWS, Red Hat, and Docker, we know the discipline needed to operate at even a 3-nines SLA, and that the current formats and packaging fragmentation won’t get us there.

“We built KitOps to solve this through open standards. KitOps is an open-source packaging and versioning solution for AI/ML projects, designed for enterprise speed, security, and consistency in cloud-native environments.” Brad Micklea, Jozu CEO

By offering a universal interchange format called a ModelKit, KitOps leverages the already adopted OCI standard (the same standard that underlies every container) to ensure seamless integration with the tools and infrastructure we already use. This enables organizations to manage, secure, and audit AI/ML projects effectively.

This is why KitOps has already been downloaded over 30,000 times in just a few months, and is being used in production by multiple organizations around the globe.

Last week it became apparent that we aren’t the only ones thinking about these issues and that several other teams want to contribute to an open solution to this universal problem. To accelerate our industry’s path to an open standard, we have submitted KitOps to the CNCF so that others can more easily contribute to it, and benefit from it.

Why KitOps in the CNCF is Best for Everyone

KitOps is built upon the Open Container Initiative (OCI) standard, which is the standard behind containers and central to many CNCF projects such as Kubernetes, container runtimes, and registries. By extending OCI principles to the packaging and versioning of AI/ML projects, KitOps creates an open standard for AI/ML project packaging that leverages an existing standard (so we don’t become an xkcd comic.)

Contributing KitOps to the CNCF will benefit everyone:

  • Promoting Open Standards in AI/ML: As AI becomes an enduring component of cloud-native applications, there’s a critical need for standardized tools and infrastructure. CNCF’s vendor-neutral stewardship ensures that KitOps can contribute to, and benefit from, a governance model that prioritizes open standards and interoperability.

  • Fostering Collaboration and Innovation: Hosting KitOps within the CNCF provides a platform for collaboration with a vibrant community of developers, end-users, and vendors. This collaborative environment accelerates innovation and the adoption of best practices.

  • Leveraging the Cloud-Native Ecosystem: With OCI at its core, KitOps naturally integrates with nearly every modern tool, platform, and infrastructure - enabling seamless deployment and management of AI/ML workloads in cloud-native environments.

  • Addressing Industry Challenges Together: The tight coupling of AI models with datasets and dependencies presents unique challenges. By collaborating with experts in the CNCF ecosystem, KitOps can help develop solutions that address these complexities, benefiting everyone using open source models, platforms, and infrastructure.

  • Alignment with an Open Mission: KitOps shares CNCF’s goal of making cloud-native computing universal and sustainable. By standardizing AI/ML packaging, we aim to reduce fragmentation and promote interoperability across the cloud-native landscape.

What’s Not Changing

From the start KitOps was built to work across any infrastructure and tool - that won’t change. KitOps will work with open source projects like Kubernetes, MLFlow, Harbor, Quay, and others, but it will continue to work with proprietary solutions from vendors like Weights & Biases, AWS, Microsoft, and others.

If you have any questions or would like to get involved with KitOps, we encourage you to join our community Discord channel, where you can engage with our team and other community members.

As always, thank you for your support!

Jesse,
KitOps Project Lead

Share this post