IML4E has reached its third milestone
With the completion of the deliverable IML4E-D4.2-Initial MLOps methodology and the architecture of the IML4E framework, the IML4E project has reached its third milestone. Accomplishing this milestone, we have defined an initial version of the IML4E framework. This includes methods, techniques and tools for efficient data preparation as well as for automation of activities in training, testing and deployment of models.
The techniques and tools for efficient data preparation are described in our deliverable IML4E-D2.3-V1.0-First version of tools for data collection, processing, and valorisation. They support data engineers by automated data error detection in tabular data, data quality dashbording, data quality estimation, data cleaning and a privacy friendly workflow for medical image processing. Information on our techniques and tools for automating training, testing, and model deployment activities can be found in Deliverable IML4E-D3.3-V1.0-First version of tools for advanced model engineering. These techniques and tools support software developers and data engineers in testing, documentation, model deployment, and systematic quality assurance and monitoring of ML-based applications. Deliverable IML4E-D4.2-Initial MLOps methodology and the architecture of the IML4E framework describes a first version of the IML4E framework as well as the IML4E OSS platform as the technical basis for the integration of the techniques and tools developed in the project. Besides others, this includes a reference architecture for an MLOps stack, a maturity assessment approach for MLOps and an initial methodology for an automated and continuous audit-based quality assurance in MLOps.
