Recently I had an opportunity to sit down with Steve Michelotti, Program Manager on the AzureGov team and talk about a Machine Learning (ML) application we built for a federal agency. This application is a great example of how AIS leverages the latest innovations on the AzureGov platform to build applications that align with agencies’ missions – and go beyond IT support to directly assist in meeting the mission objectives.

Specifically, this application was designed to help analysts get personalized recommendations (based on their own preference settings, ratings provided by their co-workers) for stories they need to analyze as part of their daily work.

Brent Wodicka from AIS described this application in an earlier blog post.

A few key points from my conversation with Steve:

1. We were able to refactor the solution to run completely in AzureGov by leveraging ML Server and Cosmos DB.
2. ML Server provides a robust environment for executing R- and Python-based algorithms. It comes with a number of built-in algorithms. Additionally, customers can bring popular open-source algorithms into the mix. We used an open-source topic modeling library for our application.
3. This solution represents an interesting hybrid of IaaS (ML Server running on a AzureGov VM) and PaaS services.

My full discussion with Steve is below. I also want to thank the AIS team members who worked hard work to build this application including Brent Wodicka, Jim Strang, Nisha Patel, and Nicholas Mark.