Continuous and Efficient Evolution of ML Systems: an Ecosystem-driven Approach. MLEvol

MLAssetSelection Tool

The MLAssetSelection Tool is a web application that helps ML engineers identify and select suitable ML assets (such as models, datasets, and services) for integration into ML-based systems. It builds on a unified knowledge graph that represents the ML ecosystem together with relevant software engineering concepts, enabling asset discovery and comparison based on both technical properties and system-level quality concerns.

By combining structured metadata with domain knowledge, the tool supports:

  • Search and filtering of ML assets using multiple criteria (e.g., task, domain, quality attributes).
  • Comparison of alternative assets in terms of suitability for a given MLS context.
  • Informed selection decisions aligned with the evolution goals of the target system.

Team: Alexandra González, Oscar Cerezo, Xavier Franch, Silverio Martínez-Fernández

Access to Tool