Bodhisattva - Rapid Deployment of AI on Containers

Abstract

Cloud-based machine learning is becoming increasingly important in all verticals of the industry as all organizations want to leverage ML and AI to solve real-world problems of emerging markets. But, incorporating these services into business solutions is a goliath task, mainly due to the sheer effort necessary to go from development to deployment. We present a novel idea that enables users to easily specify, create, train and rapidly deploy machine learning models through a scalable Machine-Learning-as-a-Service (MLaaS) offering. The MLaaS is provided as an end-to-end microservice suite in a container-based PaaS environment for web applications on the cloud. Our implementation provides an intuitive web-based GUI for tenants to consume these services in a few quick steps. The utility of our service is demonstrated by training ML models for various use cases and comparing them on factors like time-to-deploy, resource usage and training metrics.