System deployment of machine learning is one of the most sophisticated problems in the era of AI and MLOps. Based on an open-sourced document of design patterns powered by Mercari Inc., this talk will introduce practical system designs for applicable scenarios and respectively packages.
System deployment of machine learning is one of the most sophisticated problems in the era of AI and MLOps. Based on an open-sourced document of design patterns powered by Mercari Inc., this talk will introduce practical system designs, including synchronous mode, asynchronous mode, and microservice architecture. We will take a Chatbot for customer service in e-commerce as an example to go through how open sourced packages play their roles on the system architecture.
About Toby Liu
Toby Liu is a NLP researcher. He works at Mercari Inc. as a Machine Learning engineer based in Tokyo currently. He used to be a OSS community member at MozTW, COSCUP, and g0v. One of D4SG projects he participated was Predictive Risk Map of Fire Hazards in Kaohsiung 2.0. The result was presented in DataSci.tw '17, MOPCON '17, South Taiwan Statistics Conf ‘17, and PanKnowledge Festival ‘17. His research thesis in emotion analysis was published on EMNLP '18.