Master thesis: MLOps Framework for Federated Machine Learning

Master thesis: MLOps Framework for Federated Machine Learning 

The Volvo Group is one of the world’s leading manufacturers of trucks, buses, construction equipment, and marine and industrial engines under the leading brands Volvo, Renault Trucks, Mack, UD Trucks, Eicher, SDLG, Terex Trucks, Prevost, Nova Bus, UD Bus, Dongfeng, Sunwin Bus and Volvo Penta.  

Volvo Group Trucks Technology encompasses the production of state-of-the-art products for the truck brands of the Volvo Group, as well as Volvo Group engines and transmissions, through an international world class industrial environment.  

With Volvo Group Trucks Technology, you will be part of a global and diverse team of highly skilled professionals working with energy, passion and respect for the individual to become the world leader in sustainable transport solutions.  

Background of thesis project 

With the growing emphasis on data privacy, edge computing, and AI, Federated Learning (FL) has emerged as a transformative paradigm. FL enables multiple clients (e.g., devices, organizations, or edge nodes) to train a shared model without centralizing sensitive data. However, deploying, managing, and scaling federated learning systems introduces new challenges in Machine Learning Operations (MLOps) — including distributed orchestration, experiment tracking, monitoring, reproducibility, and continuous delivery across decentralized infrastructures. 

This master’s thesis aims to design and implement an MLOps pipeline tailored for federated learning environments — enabling automated, secure, and scalable model training, deployment, and lifecycle management. 

Suitable background:

Distributed Computing & Systems, Machine Learning principles 

Description of thesis work 

The aim of this project is to investigate challenges and best practices in MLOps and federated learning. The work will include the design of an end-to-end MLOps architecture that supports federated model training, orchestration, evaluation, and deployment. 

Possible research directions and objectives 

Implement tools for model versioning, experiment tracking, and automated aggregation in a federated setting. 

Integrate monitoring and evaluation mechanisms for decentralized environments. 

Benchmark the developed system against standard centralized MLOps workflows. 

Another interesting approach would be to investigate whether knowledge graphs could be used to catch concept drift in a federated MLOps scenario for vehicular testing. The research could also dive into the embedded data analytics inherent in edge computing and FL applications. 

Methodology 

Study of relevant theories and principles, and existing research papers, related to edge computing and federated learning. 

Simulate a federated learning environment using Raspberry Pis (or similar) as edge devices. 

Study and employ federated MLOps concepts with a possible research focus on embedded analytics. 

Thesis Level: Master