Brought to you by NumFOCUS Foundation and O’Reilly Media
The official Jupyter Conference
Aug 21-22, 2018: Training
Aug 22-24, 2018: Tutorials & Conference
New York, NY

Jupyter[Hub] in the corporate

Who is this presentation for?

Architects/Decision makers/Product Managers

Prerequisite knowledge

Attendees should be able to understand without any pre-requisite knowledge, though some basic understanding of Python and frameworks such as Apache Spark would be useful.

What you'll learn

Attendees will learn that Jupyter in a large corporate is possible, though requires some upfront thought and development. Attendees will learn some basics about why a traditional bank needs these tools. They will gain insight into the hurdles they will face, and how to overcome them.


Demo – Option 1 – Focus on showing an implemented solution:

This would be a demo showing some of the data mining we have done on public data sets. This may allow us to open source an example notebook or two that we have created so that attendees can walk away and play with the data and solution themselves. Internally we mix these public data sets with our proprietary data which I could talk about, though not show. This is an interesting use case in my opinion because it shows the shift in banking technology culture. Historically, banks owned and operated only on their private data, but that is gradually shifting. Recent regulations push more data into the public view which opens up competition between banks and FinTech more directly which requires us to be nimble while continuing to satisfy regulations and safeguarding our reputation. Specific parts of the demo:

• Show where we get the data from and ingest it into our platform
• Show how we visualize and inspect the data (fun charts!)
• Show how we do some analysis on the data, such as using scipy for regression and classification to predict future performance of loans.
• Some background into why this approach and technology is financially advantageous to us.

In this option we would do a quick example of the custom extension mentioned in Option 2, but not go through each of the ones below.

Demo – Option 2 – Show the customizations we implemented:

We would still show the demo from option 1 but go through it very briefly and then jump into some of the practicalities we faced when using Jupyter in house.

• Templating. Why did we need custom templates and how did we implement it?
• Custom extensions. Implementing a custom python package and auto-loading it into Jupyter instances from Jupyter (based on user’s LDAP permissions).
o This is what we did for integrating our financial analytics models into the environment.
• Authentication/Authorization. How did we integrate into our LDAP implementation?
o We need to hit multiple services instead of hitting one LDAP server.