ParallelM, a provider of MLOps, is integrating with H2O open source to drive the adoption of AI across industries. The integrations will allow H2O.ai customers to quickly deploy and manage models in ParallelM MCenter and manage ongoing lifecycle needs like model health monitoring and model retraining for models running in production.
“Part of our mission towards democratizing AI is to empower the ecosystem. As the H2O community continues to expand and grow, we are pleased that ParallelM is investing in H2O integration," said Vinod Iyengar, head of data science transformation at H2O.ai. "ParallelM provides advanced capabilities for model health monitoring and governance for all of the H2O community of customers."
The initial integration, which is available now, makes it easy to import H2O models into MCenter in either the POJO or MOJO formats. Models can then be run in batch or real-time modes. Real-time models can take advantage of the new MCenter REST endpoint, a scalable REST interface for production ML applications.
ParallelM also plans to integrate H2O models with instrumentation for the ParallelM MLOps API to generate model statistics without data scientists or even data analysts needing to touch the model code.
This is particularly important for continuing democratization of AI as data analysts are creating more and more models using automatic machine learning systems like H2O Driverless AI.
ParallelM is also adding H2O as a natively supported engine for model retraining. With this integration, MCenter will be able to automatically trigger automated machine learning platforms like H2O AutoML and H2O DriverlessAI to retrain models. MCenter will then automatically import and deploy the new version without interrupting real-time serving or batch processing or requiring data scientists to be involved in the retraining process.
For more information about this news, visit www.parallelm.com.