State of the Art Practices for Machine Learning Operations in Google Cloud

https://drive.google.com/file/d/1O5sMKXoENbimVSD4M1C_UrhrSPhosxVu/view?usp=drive_link

This thesis focuses on the topic of MLOps, specifically on the exploration and analysis of best practices for its successful implementation. A comprehensive review of the current literature in the field of MLOps was conducted to identify best practices for the design, development, testing, and deployment of machine learning models. The importance of effective collaboration between data science, software engineering, and IT operations teams for successful MLOps implementation is discussed. Based on the analysis results, a series of practical recommendations are presented for MLOps practitioners seeking to improve their ability to design and maintain scalable and robust machine learning systems. In addition, a case study is included in which a machine learning system is redesigned using MLOps best practices, thus achieving a more scalable and understandable system while reducing the costs associated with production and maintenance. This system is based on a product that should produce analytics such as sentiment prediction on a series of texts given as input related to the ESG metrics used by companies and governments today. It is hoped that this work will contribute to developing a deeper and more comprehensive understanding of best practices for MLOps implementation.