MLaaS

A mock implementation of "Stripe for Machine Learning". One can also refer to this as Machine-Learning-as-a-Service, or MLaaS.

Introduction

This is a sample code repository I use to show to potential clients. Through this project, I want the client to see

• how I write my code, including code format, comments, and git workflow,
• how I reason about, and propose solutions to, under-specified business problems,
• how I architect my code,
• my understanding of web development, and best practices.

I use the following toolset:

• Ruby on Rails framework,
• PostgreSQL database,
• Sidekiq background processing,
• Twitter Bootstrap HTML/CSS framework,
• RSpec and Capybara testing,
• Heroku hosting.

The client and the idea are something I made up for demonstration purposes.

Problem

Suppose that a client approaches me with the idea that they want to be "Stripe for Machine Learning". That is, they want to be an API platform that allows developers to use the power of machine learning algorithms without having to implement them. Incidentally, Wise looks to be doing something along those lines.

The client is looking for a proof-of-concept, also known as an Minimum Viable Product (MVP).

Solution

On a high-level, my solution is as follows:

1. Write code to define a classification algorithm. Since this is an MVP, one is enough. A simple linear regression algorithm will do.
2. Design and build out the API that a. allows the developer to post their data for asynchronous processing, b. triggers a background job to run the classification algorithm on the data, c. allows the developer to check the status of the job. If the job is done, a result is displayed. d. Design a front-end app to demo the product.

I'm going to use GitHub Issues to architect my solution incrementally. In general, a feature maps one-to-one to a Github Issue. In some cases, big features with several moving parts (e.g. front-end and back-end changes) wil be split into several smaller GitHub issues.