EscrowGPT

Inspiration

Imagine missing out on the latest restocking of Taylor Swift's limited-edition signed copy of Midnights. You quickly scour the internet for a reseller, and finding one on eBay, you quickly snag a copy. The vinyl arrives in 3 days!

Unfortunately for you, when you take the shipped vinyl to an authenticator, you realize in-person that "Taylor Swift's" signature is a fake.

For traditional marketplace websites and banking interactions, there are avenues to recoup your money by raising the issue with that website or bank. However, when exchanging items using cryptocurrency, cryptocurrency's decentralized approach makes this recuperation difficult, if not impossible to accomplish. Additionally, the lack of widespread adoption of blockchain in supply chains means that there is often no clear record of the item before it is sold using cryptocurrency. Using humans to validate the authenticity of a good is also expensive and time-consuming.

What it does

EscrowGPT is a proof of concept for an AI-assisted method of item verification. When selling or buying items on EscrowGPT, AI is used to flag items that are deemed suspicious for human validation. By using AI to flag suspicious items, human validators can spend time on cases that are of high priority. This is a huge time saver, as it can take a human a few minutes up to several hours or days to verify the authenticity of items.

How we built it

The front-end is built using React.js, while the backend uses TypeScript and Node.js. We also use the NEAR and Jackal libraries for the cryptocurrency platform and the database storage.

Challenges we ran into

Our project had a delayed start due to a variety of factors that were out of our control. Additionally, our UI/UX Designer only had beginner front-end experience, so creating a detailed front-end was difficult within the timeframe as our other teammate was focused on the backend. Although we initially opted to use Locofy to convert Figma to React, misunderstandings in the project specifications led to the creation of a simplified front-end. Finally, the lack of documentation for Jackal and NEAR presented difficulties in using the library.

Jackal Positives and Negatives

Positives:

1. Wide variety of capabilities
2. Extremely fast

Negatives:

1. Lack of documentation
2. Difficulties in setting up environment
3. Only works in browser
4. Requires the use of Keplr

Accomplishments that we're proud of

• Deepening our understanding and capabilities in blockchain
• Creating an app from start to finish

What we learned

• Figma plugins like Locofy and Anima can create React apps in a very short amount of time
• Through Locofy's suggestions, our UI/UX Designer learned about best practices in creating Figma prototypes
• A lot about how blockchain works
• How to integrate blockchain and artificial intelligence into a project

What's next for EscrowGPT

• Implementing more of the front-end to look closer to our Figma prototype
• Using Axelar to enable interchain purchases

Built With

• figma
• jackal
• near
• node.js
• react
• typescript