The open source software platform MLflow, which was first built by Databricks, and JFrog Artifactory have announced a new machine learning (ML) lifecycle integration. JFrog Ltd. (JFrog) (Nasdaq: FROG), the liquid software company and makers of the JFrog Software Supply Chain Platform, made this announcement today. With the announcement of native integrations with Qwak and Amazon (NASDAQ:AMZN) SageMaker earlier this year, JFrog expands their range of universal AI solutions by providing businesses with a single system of record that uses Artifactory as a model registry. With the help of the new integration, JFrog users now have a more efficient, end-to-end DevSecOps process for developing, managing, and delivering ML models and apps powered by generative artificial intelligence (GenAI) in addition to all other software development components. Businesses can verify the security and provenance of ML models and enable responsible AI practices by making each model traceable and immutable.
According to industry studies, at least 80% of machine learning models developed to generate new AI-powered applications never make it to market, mostly because of technical difficulties integrating the model with already-running businesses. By combining the well-liked open source model development tool MLflow with an organization’s established DevOps workflows, JFrog’s integration with MLflow helps them overcome this by providing end-to-end visibility, automation, control, and traceability of ML models from experimentation to production.
Developers and data science teams need to handle models with trust, just like they handle any software package, in order for enterprises to successfully adopt and deliver AI and GenAI-powered apps at scale, according to Yoav Landman, CTO of JFrog. This can only be accomplished with our new connection with MLflow, which offers versioning, lifecycle, and security controls in a universal, scalable, one system of record for all binaries.
JFrog MLOps: One source of accuracy for every model
The combination of JFrog Artifactory and MLflow, building on its successful interfaces with all major ML tools available on the market, gives ML engineers, Python, Java, and R developers the flexibility to work with their preferred tool stack, utilizing Artifactory as their gold-standard model registry. In addition to natively proxying Hugging Face, JFrog’s scalable and ubiquitous technology concurrently detects harmful models and enforces license compliance, enabling developers to constantly use available open source models. In order to maintain risk-free machine learning applications, the solution also includes the software security measures and scanners offered by the JFrog Platform.
MLSecOps: Reliable and Selected Models
Hundreds of malicious AI ML models have been found by the JFrog Security Research team on the public Hugging Face AI repository, raising the possibility of data breaches or assaults. This incident emphasizes the necessity for ongoing security awareness and preventative cyber hygiene by highlighting the possible risks present in AI-powered platforms.
Combining MLflow and JFrog Artifactory will enable users to more quickly and easily create, train, and implement models with improved security, versioning, traceability, and governance. This is achieved by using JFrog’s scanning environment to thoroughly inspect each new model that is uploaded to Hugging Face.
Read this blog post for a more thorough examination of JFrog’s integration with MLflow to enable ML and GenAI-powered app development. Developers can download the free plug-in here to experiment with these new features firsthand.
