Prospects for Machine Learning pipelines in the Italian industry with a comparative analysis and implementation of frameworks on a case study

Machine learning industry is expected to peak in the next years. Thanks to a growing availability of big data, connected devices and cloud services, and thanks to the growing democratization of these tools, companies are increasingly adopting them. This comes with an heavy burden on companies, who need to educate themselves and re-structure their IT systems to be able to produce and sustain such systems to make use of them. Which is why more and more players are developing tools allowing an easier integration and utilization of ML-pipelines for commercial use. The goal of this thesis is first to provide a general overview of machine learning industry, its trends, its general direction and what are the drivers and challenges that it is facing and will face in the near future. Secondly, the main goal is to provide an assessment and prospects for these technologies in the Italian industry in a context of modernization and integration of data in businesses. Finally, an ulterior objective is to provide a comparative analysis of the existing solutions that are currently offered by players in the industry and offer an alternative positioning to what is currently existing. Lastly the goal is to provide a use case providing a comparison of the adoption of these solutions vs an ad-hoc case built from scratch. The results show a strong consensus toward integration of ML pipelines in the Italian industry. Experts largely agree on a common set of drivers and factors that are pushing and will push, among which a necessity of introducing expertise and a structured approach while leveraging domain knowledge and expertise to produce business value from data. Furthermore, a cluster of common applications across sectors in the industry has been identified which represents an opportunity for pipeline frameworks producer to be integrated in companies. This inclusion is also driven by the fact that survey has revealed that companies have inadequate resources and infrastructure which opens the way for ML framework producers to insert themselves as solution for a quicker adoption of ML technology in the businesses. The ML framework landscape showed the general direction that these frameworks are uptaking, which is a combination of flexibility and easiness of use which answers to industry needs and stay aligned with the idea of democratization of ML solutions. Finally, the case study re-implementation of an ad-hoc pipeline proved that a faster approach and quicker experimentation of problems could lead to quicker and better insights on the issues while providing predictive models as accurate or more of an ad-hoc pipeline.