As we move into 2019, Open Data Group is seeing a very powerful secular shift towards the implementation of machine learning models, a trend recently mentioned in our blog “Kicking Off 2019: The Year of Model Operations”. In connection with these trends, Open Data Group is pleased to announce the release of FastScore 1.9, which continues to grow our model operations offering and capabilities. As a Docker based microservice approach, FastScore provides enhanced functionality to address the emerging needs of machine learning operations.
Supporting Model Operations for The Entire Model Life Cycle
In FastScore 1.9, we have added strategic new microservices to our FastScore suite.
As models migrate from development into production, many things will change: the compute environment, data, down-stream systems and more. Model code and FastScore assets, such as schemas, are promoted through phases, including Dev, Test and Production. Assets may be locked down by code management tools such as GitHub, and at each change there may be unique security and access considerations. The core new capabilities of FastScore 1.9 provide enhanced capability to manage and track these changes. New microservices include:
- Lineage – Inspired by our regulated industry customers (banking, insurance), Lineage captures metadata created during the testing and deployment process and generates relationships between the data. Teams are able to reproduce exact runs and audit historical runs.
- Access – is used to integrate with existing tooling around security management, such as LDAP or Okta, and allows groups of users to access and perform certain actions by model life cycle stage..
- Baker – Many of our customers found that implementing life cycle stages, and even model-specific operational contexts makes sense. With Baker, users can “bake” the preferred configurations directly into the FastScore Engine. Baker connects with container registries, providing convenient access and controls to the model operations team.
- Composer A combination of models that chain together or feed into other models is called an analytic workflow. Composer allows the user to create, manage, and deploy “workflows” of models: complex multi-model analytic pipelines that utilize multiple FastScore engines to execute. Composer comes complete with an easy to use GUI, automatic workflow generation, deployment and connection to orchestration systems like Kubernetes.
Additional Updates in the FastScore v1.9 Release
What’s next for FastScore…
Over the next 12 months Open Data Group will focus on continuing the expansion of our functionality that supports our enterprise clients, and to solve their common challenges. Our mission is to enable current and future customers to deploy ANY machine learning model into their existing infrastructure. Our customers are taking advantage of the cost and flexibility the cloud provides, and enjoy the freedom to choose any combination of model creation tools, languages and packages they need to create the best model for their use cases. And above all, together we are enabling critical mathematical assets to be operated safely, at scale, by traditional DevOps and engineering teams.