Engineering the Future of Software
29–31 Oct 2018: Tutorials & Conference
31 Oct–1 Nov 2018: Training
London, UK

Using Continuous Delivery with Machine Learning to Tackle Fraud

Sarah LeBlanc (ThoughtWorks), Hany Elemary (ThoughtWorks)
14:1515:05 Tuesday, 30 October 2018
Application architecture
Location: Park Suite (St. James / Regents) Level: Intermediate
Secondary topics:  Best Practice, Case Study

Who is this presentation for?

This talk is for developers and architects interested in machine learning and how to design systems promoting fast delivery of models safely into production.

Prerequisite knowledge

The only prerequisite for this talk is a basic understanding of CI/CD principles.

What you'll learn

The audience will walk away from this talk understanding: - Basic data science principles and how they can be applied to the problem of fraud detection - The benefits of a continuous delivery system for deploying trained models - A system architecture to allow data scientists to experiment with data manipulation and model creation - A system architecture for quickly deploying models into production both for decision-making and safe experimentation and observation

Description

A global financial institution recently tasked us with a challenging problem; detecting credit card application fraud. To keep up with their tricks, the detection algorithms (“models”) need to be updated continuously.

The client’s current practice of relying on a third-party vendor to manage these models is expensive and time-consuming. The longer these models go without updating, the less effective they become, losing money to undetected fraudsters.

Our task was to devise a process to improve their fraud detection with a sophisticated machine learning workflow that empowers data scientists to rapidly and iteratively design, develop, and productionalize new models.

Join us for a unique talk where Data Science meets DevOps to productionalize machine learning fraud detection models. After this talk, you will understand how data science algorithms can drive effective models, and how we are applying Continuous Delivery principles at a global corporation to create an extensible platform for more than just application fraud.

Photo of Sarah LeBlanc

Sarah LeBlanc

ThoughtWorks

Sarah LeBlanc is a Software Consultant at ThoughtWorks. She spent two years at one of our largest clients – a global financial institution – creating a platform and its first product, a new system for consumer loans. From there she started working on projects focusing on using machine learning for fraud and anomaly detection. Sarah finds the most joy in coaching and mentoring other developers and learning from her fantastic team members, but is not averse to debating coworkers (and gracefully accepting defeat in the face of sufficient evidence). However, Sarah continues to recognize the merits of coffee lids.

Photo of Hany Elemary

Hany Elemary

ThoughtWorks

Hany Elemary is a software consultant at ThoughtWorks, where he solves challenging business problems through clean, testable design and architecture. Over the past 11 years, Hany has worked on a number of different layers of the technology stack for highly trafficked applications. Most recently, he authored a video series, TDD with React and Redux in an Isomorphic Application. When he’s not chained to his computer, Hany enjoys traveling to new places and sipping on coffee with no lid, as he firmly believes the lid compromises the integrity of the coffee’s flavor profile.

Leave a Comment or Question

Help us make this conference the best it can be for you. Have questions you'd like this speaker to address? Suggestions for issues that deserve extra attention? Feedback that you'd like to share with the speaker and other attendees?

Join the conversation here (requires login)