End-to-End Artificial Intelligence Lifecycle Management

Abstract

In this digital era, companies are battling at the forefront of innovation to share the next transformative idea with the world. Many of these ideas apply Artificial Intelligence (AI) to solve important societal problems. However, companies may be facing difficulties understanding the core problem, understanding the data, preparing the data, building models, evaluating and finally deploying the AI technologies.

In this thesis, we propose an AI Ecosystem that enables end-to-end AI lifecycle management. This ecosystem enables teams to easily transition from concept to prototype and from prototype to deployment. We show three key pillars for a successful AI project: process, people, and platform. We also discuss the ethical and regulatory considerations of building AI technologies in this space. The study was performed at Boston Scientific with two use cases: the interventional cardiology team actively developing an AI solution using Intervascular Ultrasound (IVUS) images and the supply chain team exploring AI solutions for demand forecasting. We demonstrate how an AI Ecosystem can enable such teams to focus on their core responsibility, developing innovative medical solutions that improve patients lives.