Inspiration

The current United States is in an interest-rate rise environment. This begs the question "what will happen to mortgage rates in the near feature"? How will lenders address the way they do business to stay ahead of the competition and deliver value to clients.

What it does

If you know something about a house and the people who are doing your home appraisal, you can assign properties to the appraisers who tend to give fair valuations to specific types of homes. My tool lets a lender gauge how uncertain valuations will tend to be on a property and which home appraiser is the best person for the job.

How I built it

My tool leverages machine learning models trained on a publically available dataset on Kaggle. The analytics and inference capability are wrapped behind a user interface to minimize the amount of software expertise that a user needs.

Challenges I ran into

I had 6 million datapoints out of which 100,000 were useful. Even then, it was difficult to process such a large dataset in the time allowed. As a result, the machine learning models would benefit from further tuning and the UI can be developed further in terms of ease of use and features available.

Accomplishments that I'm proud of

We did everything on our own: data preprocessing, machine learning model training, UI & data management. The end result is pretty enough but furthermore demonstrates my idea end-to-end.

What I learned

It's hard to solo an end-to-end project.

What's next for Blend

See the uploaded slide deck.

Built With

Share this project:
×

Updates