Heya HN, after spending +1 year building an ML-driven analytics product (that didn't pan out unfortunately), I've pivoted to solving a problem my team and I found while building the previous product … why the hell is it so hard to move a model from a Jupyter notebook, to a development server, then to a production pipeline!?<p>To solve this my team and I started the open source KitOps project under the Apache 2 license. KitOps includes the Kit CLI that uses a Kitfile manifest to create ModelKits:<p>1. The kit CLI packages your model, datasets, code, and configuration into an OCI compliant artifact called a ModelKit. The ModelKit keeps everything in one place, tagged, and versioned, our team has found it much easier to collaborate.<p>2. ModelKits are modular (unlike a Docker file), you can pull only the model or datasets, for example, or grab the whole thing. Plus you can still create a dockerfile for your serialized model.<p>3. ModelKits provide a history of meaningful state changes for auditing and are immutable so they should be great for a secure bill-of-materials (SBOM) initiative. Full disclosure, I haven't used SBOMs myself but a friend mentioned this as a benefit so thought I'd pass it along.<p>4. The Kitfile is YAML and easy-to-read so even people who don't understand the ins-and-outs of ML development can find what they need to integrate models into their apps, test models with validation data, or deploy models to their inference engine of choice.<p>We're still early in the development process and are really interested in collecting community feedback.<p>You can learn more about KitOps at <a href="https://kitops.ml" rel="nofollow">https://kitops.ml</a><p>Or checkout the source code here: <a href="https://github.com/jozu-ai/kitops">https://github.com/jozu-ai/kitops</a>