Payeasy tackles one of the most fundamental challenges in the payments space. Over 75% of business to business transactions are processed the old fashioned way. This means that when a customer purchases a product, the merchant emails them a PDF Invoice of the purchase. The customer than forwards this bill to their accounts payable division, who spends hours manually entering data from the bill and categorizing it for the banking and accounting systems. The customer then mails a check to the merchant, which can take weeks for the merchant to process. We found the billpay process to be ridiculously inefficient, and sought to build a hack to fix it.

What it does

Payeasy utilizes PDF parsing and machine learning algorithms to automate the data entry and categorization of PDF invoices. We make billpay as simple as uploading the PDF of a bill. Our system automatically extracts the important data from bills that customers upload, and categorizes them. Customers are also able to rate and review their vendors, essentially making this platform comparable to a B2B version of Yelp. Backed by the Capital One API, all bills entered into our system are automatically synced with customer banking and accounting systems. Moreover, we use the Square API to make paying these bills a one-click solution.

How we built it

We built our frontend in React JS and our backend in Python/Django. We also developed optical character recognition algorithms to extract the data from the invoice PDFs. Furthermore, we used machine learning for the categorization of the bills.

Challenges we ran into

Some of the main challenges we ran into included the accuracy of the PDF parsing and the optical character recognition.

What we learned

We learned about the versatility of the APIs we used (such as Capital One's API) and their wide variety of use cases. We also learned more about parsing algorithm optimization, how invoices work, and sci-kit-learn for machine learning.

What's next for PayEasy

Moving forward, we plan on allowing users to query over all merchants in the system, allowing to be able to select the vendors of their choice based on prior ratings and reviews. We also plan on training neural networks to track, classify, and understand customer spending, vendor reviews, etc. PayEasy has the potential to become the hub for B2B transactions and business relationships.

Share this project: