MLOps Your Way with the JFrog Platform
New Machine Learning repository, FrogML SDK, and JFrog ML accelerate delivery of trusted AI/ML applications
Just like in traditional software development, creating AI applications isn’t a one size fits all approach. However, many of the challenges and concerns facing AI/ML development teams share common threads – difficulties getting models to production, tangled infrastructure, data quality, security issues, and so on.
Regardless of how you build it, to accelerate production-ready AI, AI/ML workflows must be seamlessly connected to an organization’s standard DevOps and Security frameworks and practices, operating from a single source of truth across MLOps and DevSecOps.
With the JFrog Platform’s new advanced model registry capabilities paired with JFrog ML, organizations can diminish infrastructure challenges and apply the same mature controls that accelerate their traditional software development to AI development.
New Machine Learning Repo + FrogML SDK
Uniting the AI/ML and traditional software supply chain starts by operating from a single source of truth. Your AI/ML artifacts should live alongside your traditional software artifacts. The standard practices your DevOps team have implemented for building and deploying traditional software applications should be applied to AI/ML development and deployment.
JFrog Artifactory’s new Machine Learning Repository, paired with the new free FrogML SDK, enables organizations to take a giant step forward in doing just that. With these two tools, you can easily bring JFrog into your AI/ML workflows with zero disruption for Data Science and AI developers.
The new Machine Learning repo type visible in the the repository creation flow
The FrogML SDK is a smart and optimized Python library providing advanced ML model management capabilities by seamlessly integrating JFrog Artifactory as the primary model store. The FrogML SDK will also become the primary library that is utilized for JFrog ML, allowing you to take advantage of it for ML Models, Tags, Deployments, Batch Executions, etc.
By opening up the JFrog Platform to your data science teams, trusted components are pulled from Artifactory (python packages, containers, etc.) and used as they construct their models. Leveraging the FrogML SDK, model artifacts that come out of the experimentation process are stored in Artifactory’s Machine Learning repository where they are versioned, managed, secured and easily accessible by the appropriate dev teams. Customizable properties allow for the tagging of models with things that make sense to developers, making them easily searchable and identifiable.
Installing FrogML in a Jupyter Notebook and pushing artifacts to Machine Learning repo
FrogML and Artifactory’s Machine Learning Repository are context aware, allowing teams to work in the way that suits them best while enabling the capture of rich metadata to enhance what’s known about the model. Today, it supports the following flavors:
- Catboost
- HuggingFace
- ONNX
- Scikit-learn
- PyTorch
- Generic Python function
And because the SDK is lightweight, your data scientists have adopted your DevSecOps best practices with literally just a couple lines of code, delivered through the JFrog Platform.
Expect to hear more about the FrogML SDK as we continue to enhance its capabilities over the coming quarters.
Bring your own toolchain or step up to an all-in-one Enterprise ML Solution.
Staying true to our universal roots, the JFrog Platform allows your data science teams to work the way they prefer to work. In fact, we already have out-of-the-box integrations with SageMaker, MLFlow, and NVidia. Your AI/ML teams can continue to use your proprietary, open-source, or third-party tools, but by using JFrog’s Machine Learning repo and FrogML SDK you can start to apply your existing enterprise development standards and processes to AI development.
Alternatively, you can ditch the tangled web of DIY ML stacks for JFrog ML, an all-in-one approach to build, deploy, manage and monitor all your AI/ML workflows.
JFrog ML overview dashboard
JFrog ML brings together the tools, integrations, environments, and out-of-the box approach needed for successful AI/ML development. Data Scientists and ML Engineers benefit from simplified testing and experimentation, and not having to manage infrastructure, allowing them to focus on innovation. Crucially, JFrog ML takes care of the actual deployment and serving of the model, a final hurdle that trips up so many teams in actually bringing a model service to production.
Attend the upcoming group demo of JFrog MLInstill trust into your AI/ML workflows
As organizations turn to AI to bring new value to the market and outpace the competition, it’s critical to bring its development in line with traditional enterprise development practices. The JFrog Platform is the only solution that unites DevOps, DevSecOps, and MLOps so that organizations can move into the AI future confidently with trust in the inputs, workflows, and outputs of AI development. See for yourself and start a free trial of the JFrog Platform today.