Automation and Orchestration for Machine Learning Pipelines - A Study of Machine Learning Scaling - Exploring Micro-service architecture with Kubernetes

Aug 3, 2026

Upload by Miles

Machine Learning (ML) has rapidly expanded beyond algorithms and coding, creating strong needs for building and managing complete ML projects. This thesis investigates how a minimal engineering team can create and maintain an MLOps pipeline. The work explores containerization and micro-service orchestration, producing a minimal on-premises Kubernetes cluster on physical servers and Ubuntu VMs. The cluster scales by adding nodes, uses a locally shared network folder as external storage, and supports local or cloud container registries. After setup, the system trains YOLOv5 on a custom dataset and later performs distributed DDP training with PyTorch, TorchX, PyTorch Lightning, and Volcano.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages78

Reading Time3 h 17 min

AccessFree to download

Contents

Introduction

Background

Methods

Results

List of Abbreviations

Automation and Orchestration for Machine Learning Pipelines A Study of Machine Learning Scaling: Exploring Micro-service architecture with Kubernetes Master’s thesis in Complex Adaptive Systems FILIP MELBERG VASILIKI KOSTARA DEPARTMENT OF PHYSICS CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2024 [www.chalmers.se](www.chalmers.se) Master’s thesis 2024 Automation and Orchestration for Machine Learning Pipelines A Study of Machine Learning Scaling: Exploring Micro-service architecture with Kubernetes FILIP MELBERG VASILIKI KOSTARA Department of Physics Chalmers University of Technology Gothenburg, Sweden 2024 Automation and Orchestration for Machine Learning Pipelines A Study of Machine Learning Scaling: Exploring Micro-service architecture with Kubernetes FILIP MELBERG, VASILIKI KOSTARA © FILIP MELBERG, VASILIKI KOSTARA, 2024 . Supervisors: Hamid Ebadi, Infotiv Technology Development & Giovanni Volpe, Department of Physics Examiner: Giovanni Volpe, Department of Physics Master’s thesis 2024 Department of Physics Chalmers University of Technology SE-412 96 G¨oteborg Sweden Telephone + 46 (0)31-772 1000 Chalmers digital printing Gothenburg, Sweden 2024 Automation and Orchestration for Machine Learning Pipelines A Study of Machine Learning Scaling: Exploring Micro-service architecture with Kubernetes FILIP MELBERG, VASILIKI KOSTARA Department of Physics Chalmers University of Technology Abstract Although Machine Learning (ML) has been around for many decades, its popularity has grown tremendously in recent years. Today’s requirements show a great need for the development and management of ML projects beyond algorithms and coding. The aim of this thesis is to investigate how a minimal team of engineers can create and maintain a ML pipeline. To this end, we will explore how a Machine Learning Operations (MLOps) pipeline could be created using containerization and container orchestration of micro-services. After relevant research, the result is a minimal, on-premises Kubernetes cluster set up on physical servers and Virtual Machines (VMs) running the Ubuntu Operating System (OS) . The cluster consists of a master and two worker nodes, which are used for two main ML frameworks. Populating the cluster with more nodes is straightforward, which makes scaling a simple task. Additionally, a locally shared folder on the network is mounted in the cluster as an external storage and the cluster is configured to access either a local or a cloud-provided container registry. Once the cluster is set up and running, an application is launched to train the YOLOv5 model on a custom dataset. Later, Distributed Data Parallel (DDP) training is performed on the cluster using PyTorch, TorchX, PyTorch Lightning and Volcano. Keywords: DevOps, Docker, Kubernetes, Micro-service, ML, MLOps, PyTorch, YOLO vi Acknowledgements The research for this master thesis was carried out within the SMILE IV project, financed by Vinnova, FFI, Fordonsstrategisk forskning och innovation under the grant number 2023-00789 [75] . We would like to express our deepest gratitude to our project supervisor Dr. Hamid Ebadi, Senior Researcher and Competence Leader at Infotiv Technology Development, for his invaluable guidance and his constant engagement in our thesis project. We would also wish to extend our gratefulness to Maria Kindmark Alemyr, Consultant Manager at Infotiv Technology Development, as well as the company personnel, for providing beneficial advise and access to crucial resources, such as software tools and technical infrastructure. Finally we want to thank our supervisor and examiner at Chalmers University of Technology Giovanni Volpe, Senior Lecturer at Institution of Physics at Gothenburg University, for his direction concerning administrative processes. Filip Melberg and Vasiliki Kostara, Gothenburg, June 2024 viii Contents List of Figures x List of Tables xii List of Abbreviations xiii

FAQs

What is the main goal of this thesis?

To investigate how a minimal team of engineers can create and maintain a Machine Learning Operations (MLOps) pipeline beyond just writing ML algorithms.

How is the MLOps pipeline implemented?

It uses containerization and container orchestration of micro-services, resulting in a minimal on-premises Kubernetes cluster with a master and two worker nodes.

How does the system support scaling and training workflows?

Scaling is enabled by adding more Kubernetes nodes, and the cluster mounts shared external storage and can use a local or cloud container registry. Training includes YOLOv5 and later distributed DDP training using PyTorch-related tooling.

Aug 3, 2026

Upload by Miles