Défense de mémoire de Monsieur Benjamin Jonard
Automating ML Pipelines: A MLOps-GitOps Workflow implementation
Date : 17/06/2025 09:00 - 17/06/2025 10:30
Lieu : Salle académique
Orateur(s) : Benjamin Jonard
Organisateur(s) : Benjamine Lurquin
Machine learning projects face significant deployment challenges, with only a small percentage
successfully reaching production due to operational complexity.
This thesis addresses these
challenges by developing and implementing a state-of-the-art MLOps-GitOps work#ow designed
to support machine learning projects from initiation through production maturity.
Following a comprehensive literature review of current MLOps methodologies and frame-
works, we implement a modular work#ow using a two-phase approach: initial validation with a
controlled demonstration model, followed by a design tailored for integration with the real-world
LSFB (Langue des Signes de Belgique Francophone) production system.
The implementation integrates enterprise-level technologies including Kubernetes, Docker,
Helm, Airflow, and Kubeflow pipelines, applying GitOps methodology consistently across DataOps
and MLOps processes. The resulting work#ow successfully demonstrates how a state-of-the-art
MLOps-GitOps approach can bootstrap machine learning projects and support their progres-
sion toward greater maturity, providing a reusable foundation for diverse data engineering and
Keywords:
machine learning projects across various organizational scales.
MLOps, DevOps, DataOps, GitOps, Machine Learning Deployment, CI/CD, Pipeline
Automation, Kubeflow, Airflow, Python, Kubernetes
Contact :
Benjamine Lurquin
-
081725255
-
secretariat.info@unamur.be
Télecharger :
vCal