Topic: MLOps and distributed machine learning on the edge of the network
About Me:

My name is Ari Kukkaro and I am a Ph.D. student at Tampere University. My journey towards Ph.D. started during my Master’s thesis when I worked as a research assistant in Tampere University’s PRODI project. During that project, with the help of my current supervisors Assoc. Prof. David Hästbacka and (now) Post-doc Sergio Moreschini, I developed a prototype of a vapour detection system, where machine learning tasks were distributed over edge and fog layers, but also MLOps-style continuous training pipeline was applied. After that, I continued to work on this topic.
Research focal point:
My research focuses on distributed machine learning (ML), where data-intensive, complex computational tasks are distributed across multiple devices or computing units. This type of approach reduces memory consumption and speeds up ML model training time compared to traditional centralized machine learning approaches, and in some cases even improves data privacy. I am also exploring how ML tasks can be distributed specifically across edge and fog layers. In short, edge computing may have limited computing devices, while the fog layer provides richer resources for performing computational tasks. Both computing paradigms have in common that they provide low latency services from a network perspective.
Then, to take advantage of both distributed machine learning and near-the-network computing paradigms, I explore how they can be applied to the development and deployment of agile and fast machine learning applications, MLOps.