KNIME for Continuous Deployment of Data Science
Define and automate a ModelOps process to accelerate time to insight and scale the impact of your data science.
From creation to large-scale dissemination of data-driven insights
Securely deploy data science solutions with a standardized (but fully customizable) validation and release test process.
Automate some or all of the productionization process to increase the volume and velocity of data science across your organization
Automate the monitoring and retraining of in-production models to ensure accuracy and efficacy. Keep an audit log, as well as enable rollback to previous models if needed
I think it's something that's becoming more and more important. It's not enough to build a data science use case or like a super nice machine learning model. It doesn't have any use if you can't put it into production.
Christine Barg, Data Scientist, Deutsche Telekom
Why KNIME for Continuous Deployment of Data Science
Define and automate the productionization of data science
- The Continuous Deployment of Data Science (CDDS) extension helps data and IT teams ensure only validated data science solutions are deployed. The KNIME Business Hub extension includes a set of KNIME workflows and data apps that are pre-configured but can be adjusted as required. Enterprises can add their own validation and governance requirements, use internal archival setups, and change monitoring and retraining strategies.
- CDDS leverages key enterprise features of KNIME Software such as integrated deployment, KNIME Hub spaces with defined execution contents, workflow triggers, and more. Using its intuitive UI, data experts can easily deploy workflows, validate, monitor and retrain them, while administrators can oversee the entire deployment process.
Easily prepare models for production (no rework required)
- Close the gap between data science creation and putting results into production. Using KNIME’s low-code/no-code environment, capture your model and all data pre-processing steps to be automatically integrated and ready for reuse in production.
- Synchronize workflow creation and production processes to ensure model monitoring can automatically trigger retraining and instant redeployment for optimal model performance.
Validated production processes
- Ensure compliance with corporate governance practices and allow only validated workflows to be moved into production. With KNIME’s CDDS extension, data science applications and (API) services are checked and validated as they move – automatically or manually – through spaces for each test, validation, and production stage.
- Add or adjust spaces as per an organization’s need. Administrators can oversee the production process at a glance using the intuitive admin data app.
Transparency and auditability
- Keep track of models in production for complete auditability. As model performance is continuously checked, the results are logged automatically, providing full transparency and the ability to roll back to previous versions as required.
- Implement event logging throughout the KNIME CDDS extension and can modify as required.
Automate as you want, including, retraining models
- The deployment processes vary between organizations depending on industry, organization size, IT regulations, and more. Automate as much or as little as you want.
- Move workflows manually through stages via intuitive data apps or fully automate validation and production. With data changes, models’ efficacy can be impacted. CDDS allows you to continuously monitor models’ performance and automatically trigger retraining as required.
