Upcoming talk: FATES properties in MLOps (Explain’AI 2025)

🇫🇷 The FATES-MLOps consortium will present Prise en compte des propriétés FATES en MLOps: perspectives et ambitions at Explain’AI 2025, on January 28th, 2025 at INSA Strasbourg. Now in its fourth edition, the workshop is backed by the CNRS GDR RADIA and hosted by the EGC conference.

The talk sets out the ambitions of the ANR-funded half of the project for a francophone audience: what the FATES properties (Fairness, Accountability, Transparency, Ethics, Security) actually demand once you try to formalize, measure and justify them inside an MLOps pipeline, rather than merely name them.

Abstract

The talk and its abstract are in French. The following is a translation.

The MLOps movement takes up the DevOps objective of reducing the gaps between development and operations teams, by integrating collaboration with data scientist teams and the phases related to building and deploying Machine Learning (ML) models. The ANR FATES-MLOps project sets out to study extra-functional properties such as fairness, accountability, transparency and security, grouped in English under the FATES acronym. By drawing on and refining proven software engineering concepts and tools, we aim to propose a systematic, tool-supported approach for taking these fundamental properties into account throughout the lifecycle of software developed following an MLOps approach. The technological obstacles concern the formalization and measurement of these properties according to context, and their systematic handling within the MLOps process through suitable mechanisms and algorithms. This involves analysing and designing the workflows used to build the models, the integration and deployment processes, and the justification that these properties are met.

Reference

Mireille Blay-Fornarino, Jean-Michel Bruel, Sébastien Mosser, Frédéric Precioso. Prise en compte des propriétés FATES en MLOps: perspectives et ambitions. Explain’AI (CNRS) 2025. [MacSphere]