Given the rapidly changing technological landscape and the growing significance of Machine Learning (ML) across diverse industries, the adoption of guidelines specific for the development of ML-based software has become crucial. MLOps is therefore defined as a combination of ML and DevOps practices, i.e., set of principles and best practices that ensure efficient and reliable deployment, management, and maintenance of software.
Some of the benefits of following MLOps guidelines include mitigating risk, improving model performance, and promoting reproducibility and scalability. Moreover, when deploying software, MLOps play plays a crucial role in the broader context of the Cognitive Cloud Continuum Continuum. The ability to leverage the computing power of devices within the same environment could facilitate both computation and information delivery.
“In my thesis, I investigated the different concepts related to the Cognitive Cloud Continuum as well as the guidelines to develop and maintain ML-based software. In the first case it was important to shed light on a definition that has been used differently based on the scenario. In the second case it was important to provide clear examples of the difference on developing software compared to continuously delivering ML-based software and what are its risks,” says Sergio Moreschini.
“I think that, just like DevOps with continuous delivery was crucial in the past, adhering to MLOps principles will emerge as the prevailing norm in an era where machine learning applications, including other artificial intelligence related software, are being released frequently.” he says.
The research methods and results have been published in five main scientific publications: three journal articles, one workshop and one conference paper. The venues include Elsevier Journal of Systems and Software (JSS), IEEE Software, and IEEE Access.
Public defence on Friday 17 November
The doctoral dissertation of MSc. Sergio Moreschini entitled Applications of MLOps in the Cognitive Cloud Continuum will be publicly examined in the Faculty of Information Technology and Communication Sciences at 12 o'oclok on Friday, 17 November 2023. The public defence will be held at Hervanta Campus in auditorium TB109 of Tietotalo building (Korkeakoulunkatu 1, Tampere). The opponent will be Dr. Andrea Janes from FHV Vorarlberg University of Applied Sciences, Austria. The Custos will be Associate Professor (tenure track) David Hästbacka from Tampere University, Finland. The work has been co-supervised by Associate Professor Davide Taibi from Tampere University.