IN-CRE-D (Intelligent Credit Decisioning using AI)

Inspiration

The business problem of credit decisioning process for corporate customer is quite challenging leading to bad loans and poor customer experience. We wanted to address this problem using Artificial Intelligence to enable Analysts with the insights/intelligence to make informed decisions

What it does

For a Corporate customer, Solution helps the Analysts to: a) Evaluate the risk profile of customer by leveraging external sources to get risk details b) have all the regulations at their finger tips and Analysts can get exact details with an interactive agent based interface c) Extract and identify the legal clause extraction in contract documents so that it can be quickly referenced

The solution interacts with Finastra API's with Search Corporation API and Create New Corporation (for customer onboarding for new customer)

How I built it

The building blocks of the solution was meticulously designed to address all the challenges of the business problem. The functionalities were also designed to be able to get easily integrated into external environment. The solution uses Artificial Intelligence (using various Deep learning models) to help the Analysts make informed decisions

Challenges I ran into

There were several challenges including training the state of the Art deep learning models for comprehension and building deep learning models for document text inference. There were several components to the solution and making it to work in an integrated way was also a challenging aspect of the solution.

Accomplishments that I'm proud of

The overall solution shaped up pretty well in addressing all the business challenges of Credit Decisioning process.

What I learned

The complete background of business challenge helped in identifying the right technology solutions in addressing the same.

What's next for IN-CRE-D

More technology innovations in the solution in integrating to the existing Credit Decisioning process and introducing more intelligence to Analysts using Knowledge Graphs etc.