Verta.ai announces the release of ModelDB 2.0 a pioneering model versioning solution to make ML models reproducible

PALO ALTO, Calif., March 17, 2020 (GLOBE NEWSWIRE) -- Verta.ai, provider of Verta Enterprise, an open-core end-to-end MLOps platform, today announced the launch of ModelDB 2.0, an industry-leading, open-source model versioning system to make machine learning (ML) development and deployment reliable, safe, and reproducible. In a field that is rapidly evolving but lacks infrastructure to operationalize and govern models, ModelDB 2.0 provides the ability to track and version the full modeling process including the underlying data and training configurations, ensuring that teams can always go back and re-create a model, whether to remedy a production incident or to answer a regulatory query.

“In the insurance industry, many models have to be submitted to regulators for audit and documentation. However, the process for creating a model is more like research; it takes time and is ad-hoc by nature. Documenting and reproducing models is a massive undertaking for our team and actuaries spend weeks answering questions from regulators. As a result, for us and many regulated companies, making models and analyses reproducible is essential in our business,” said Samuel Madden, Chief Scientist at Cambridge Mobile Telematics, a leading InsureTech company that uses data and ML to make drivers safer.

ModelDB 2.0 pioneers model versioning akin to source version control for code

While robust systems are available in software for key operations like source code management, agile delivery, and operations; these systems are missing for ML models, making it challenging for companies to integrate ML into their core products.

ModelDB 2.0 completely reconsiders what a model versioning system should provide and how it should be built. Using the best-in-class constructs from code versioning systems like Git, and adapting them to the special reqirements for reproducing ML models, Verta’s ModelDB 2.0 allows for complete governance, audits, version control, and collaboration on ML models.

Licensed under Apache V2, ModelDB 2.0 is now generally available, delivering the following new capabilities to the open-source community:

• Ability to version the key ingredients of a model including code, data, configuration and environment
• Ability to reproduce any model that has been versioned using the ModelDB protocol
• Integrations into popular ML frameworks such as PyTorch, Tensorflow, and scikit-learn
• User management with authentication, authorization, organization and teams

ModelDB 2.0 from Verta.ai Helps Organizations Bring Agility to Data Science

The first version of ModelDB developed at MIT was focused on tracking model metadata (e.g., who created the model, the exact configuration for model training etc), and many research and commercial systems went on to adopt this approach introduced by ModelDB. However, years of experience building and deploying ML models, and the opportunity to work in regulated industries highlighted to the Verta team the inadequacies in this approach.

“While metadata is useful and can tell you things like who built a particular model, it does not provide the ability, say, to go back in time and re-create that model to answer a regulatory question. As models make key business decisions, the industry demands greater accountability and safety on how our models are created and used”, said Manasi Vartak, Verta.ai CEO and creator of the original ModelDB project at MIT.

Model versioning provides a robust foundation for downstream MLOps like packaging, deployment, operations, and monitoring, and Verta has seen the benefits of grounding Verta Enterprise in a robust model versioning system of ModelDB.

Learn More

• ModelDB 2.0 is released under the Apache V2 license and is available today at: https://github.com/VertaAI/modeldb.
• Get started with ModelDB 2.0 with the release blog post here and associated tutorials.
• Tune in for a ModelDB V2 Release Webinar on April 1st at 10 am PST.

About Verta.ai

Verta.ai is a Palo Alto-based developer of Machine Learning Infrastructure Software. The open-core Verta Enterprise Platform provides end-to-end MLOps capabilities to Data Science and ML teams to enable them to bring more models into production rapidly and safely.

Verta.ai started from the ModelDB research project at MIT. Developed by the Verta team at MIT, ModelDB was the first open-source model management system and went on to get significant traction in financial services and research labs alike. Verta.ai has continued to expand its offering from the ModelDB system to the complete Verta Enterprise MLOps platform.

Press Contact

Julia Mohler, Director of Communications for Verta.ai at Formulated.by
E: [email protected] and [email protected]

[ Back To TMCnet.com's Homepage ]