search menu icon-carat-right cmu-wordmark

Marvin AI: An Open Source Platform to Deploy and Manage Machine Learning Models

May 2018 Presentation
Daniel Takabayashi (B2W Digital), Jeremy Elster (B2W Digital)

Marvin AI is an open source project that helps data science teams deliver complex solutions with a high-scale, low-latency, language-agnostic, standardized architecture.

Publisher:

Software Engineering Institute

Abstract

The recently open-sourced platform called Marvin AI is helping B2W Digital data science teams to put some interesting projects into production, such as the dynamic pricing and fraud detection. The platform is helping B2W avoid challenges that are common in machine learning projects, like building a hypothesis prototype and putting it in production, enabling it to handle ever-increasing data sets, and supporting the necessary throughput of the largest e-commerce company in South America.

Marvin AI is made up of tools, libraries, an embedded server that exposes microservices to serve and manage models, and a very useful design pattern, DASFE, that simplifies the process of building new machine learning engines. In this 90-minute hands-on tutorial, one of the project architects and main open source contributor will explain all the core features of Marvin.