Machine Learning Ops (MLOps) is a set of practices inspired by DevOps that supports the creation of automated pipelines that ensure the continuous delivery and training of machine learning (ML) systems.
Make sure your data scientists are focused on solving business problems and less concerned with infrastructure data movement, and less reliant on slow processes for uploading models into production.
Use the marketplace concept to democratize data science models and training features.
Obtain improvements in the entire machine learning life cycle:
- Automation. With the adoption of MLOps techniques, it is possible to define and automate processes, standardizing and streamlining the development cycle of ML systems.
- Continuous processes. Standardization of integration, delivery, training, and model monitoring processes.
- Versioning. Efficient versioning of experiments, creation and delivery of models in production, making the follow-up on stages of development easier and helping the creation of versions of machine learning models.
- Replication. Adopting design pipelines, creation, and delivery will make the project replication process more efficient.