Personalized Lending

Inspiration

When considering a mortgage, it's typical for a person to start by asking friends and family for advice on what mortgage program they should chose. Our solution is similar to that, but better. Friends and family don't have the same financial situation, age, dependents, live in the same area, etc.

What it does

With a few pieces of information about the consumer, you can call our open API to be hosted on the FFDC platform and we will tell you what type of mortgage other people in their situation have chosen. The decision for what mortgage program best fits you comes from machine learning model.

How we built it

Using Azure Automated Machine Learning, we created a series of ML models and deployed the one with the best accuracy. We leveraged our existing ETL process for Fusion Mortgagebot LOS to get a cleaned data set consisting of 300,000 closed loans from the last year. Machine Learning models are already being built off this process which will significantly reduce our time to market.

Challenges we ran into

No one on our team is a data scientist or machine learning experts. Given access to the innovation team and their data scientists we could considerably improve models accuracy.

Accomplishments that we're proud of

We are proud that we created a fully functional prototype leveraging new technology. Our open API is product agnostic and ready to be hosted in FFDC. The machine learning model we created poses a 62% accuracy which can greatly be improved upon and lends credibility to our idea.

What's next for Personalized Lending

Mortgages are just the beginning. We want to create a suite of APIs you can use to personalize the entire banking experience for your consumers. Should they open a savings account, money market, or CD? Should they refinance their house? Should they get a personal unsecured loan or a home equity loan? Should they buy a home or continue to rent? Our objective is to Personalize lending today, Personalize finance tomorrow.