AI-Powered Mortgage Advisor

Problem Our Hackathon Project Aims to Solve:

Imagine you are a mortgage broker using ExpertPro (a Filogix's mortgage application solution), about to submit a loan application to a lender on behalf of your client.

But you're unsure if the loan request will be accepted or denied if you go ahead and submit your client’s application.

Repeatedly sending applications is taxing for the system to process, and it would be time consuming for the lender to keep re-evaluating your applications.

This scenario inspired our hackathon team to create a proof of concept for an AI-powered mortgage advisor, with the goal of advising mortgage brokers the chances of their mortgage application being accepted or denied by a lender, which will provide the mortgage applicants a faster turn-around for their application hence improve the efficiency of application process.

What Our Project Does:

By leveraging data driven, machine learning technology, we aim to demonstrate a Proof of Concept to predict the probability of the application being approved by a Lender, via a simple button click from a Mortgage application system, for example, Filogix’s ExpertPro. This change will bring at least 2 major benefits to existing mortgage application solutions: 1.Improve the efficiency of application process to better serve the applicant; 2. A faster turn-around for Applicant’s application.

Technology Used:

Machine-learning, Neural Network, Flask, Keras, Python, Tensorflow, Vue.js

Accomplishments that We're Proud of:
--From technology perspective, we have managed to train the model to output a “yes” or “no” with 91% accuracy;

--From business perspective, the quoted below are some examples of what we've received:

Feedbacks and Comments From Business People:

Barb Stanford --Lead Product Manager

"This is something that the Cdn Mortgage Brokers and Lenders would value. Great demo, very clear language and the idea is very applicable!"

John Rodrigues --Client Support Lead

“This is great! Faster turnaround time for the applicant and fewer submissions to the lender that they will end up declining.” 

Richard Audette --Senior Manager, Channel Development, Sigma Loyalty Group

"This is something I always wanted to do - take historical mortgage decisioning data and feed it into a model. I think it would also be interesting use this data to build a model to recommend a lender suitable for a given application."

What We Learned
In addition to Machine Learning expertise, we have learned, by our personal experience, the spirit of Hackathon: the passion for new ideas, new technologies; determination and teamwork.