NovaTax

Inspiration

In many online marketplaces, products are frequently mislabeled or misclassified under incorrect sales tax codes. These inaccuracies can lead to significant financial penalties for the hosting platforms (such as Walmart or eBay), as well as compliance challenges for vendors. NovaTax aims to bridge the gap between vendors and marketplaces by providing a streamlined, accurate system for product tax classification, helping both parties maintain compliance and reduce the risk of costly tax errors.

What it does

NovaTax automates the classification of retail products into the appropriate sales tax codes. The system collects information about a product either entered manually by a user or extracted automatically from product data and images. Then it compares it against a comprehensive database of tax codes and their defining characteristics. It then calculates a confidence score to indicate how well a product matches a particular tax code.

How we built it

We built NovaTax as a web-based app integrating several components. For our database we used PostgreSQL. We also used Django rest framework for the back end to use the API's. To bridge the front end and back end we used RESTFUL API. For the front end we used Next.js.

Challenges we ran into

One of our main challenges was finding and integrating a suitable tax code API that contained detailed, structured datasets for our use case. Many APIs lacked the necessary granularity or consistency, so we iterated through several before identifying one that aligned with our classification logic. Additionally, ensuring consistent data parsing, matching accuracy, and performance across diverse product inputs required extensive testing and refinement.

Accomplishments

We are proud to have built a fully functional prototype capable of accurately matching products to tax codes and generating confidence-based classifications. For a team new to large-scale data integration and compliance systems, achieving a working backend and end-to-end pipeline was a major milestone.

What we learned

We learned about the importance of logic when integrating AI into any program. We also learned how difficult it can be to accurately categorize products.