Recoop

hero section
analytics page of the dashboard
table management for the refunds
real time call automation

Inspiration

When looking at airlines we saw one financial issue that plagues every human at one point. The frustration of having your flight cancelled and having to call customer support for a refund, or even worse when you buy a ticket and it drops down in price the next day. We sought to fix that issue by letting generative AI take the wheel of both calling customer support and keeping track of price changes throughout the days.

What it does

Recoop takes in tickets or items that could fluctuate in price and processes the information continuously looking for the moment it drops or shipment cancels and it grabs the customer support number associated with the item and calls them. This process is something you can look like and is what is broadcasted in the dashboard if it successfully went through or not.

How we built it

We were able to build the front end with react, the back end is done in FastAPI and SQLite with Clerk for authentication. For AI we use Gemini for dialogue and ElevenLabs for text to speech and speech to text. For calling customer support we used bland.ai which handles our real-time negotiation. As for animations and visuals we mostly used Figma and had to use their own AI to parse the design and give us working React code, which we then expanded on and hooked up to a lot of our back end.

Challenges we ran into

We have to connect a lot of APIs together to make things happen streamline. Especially with having calls go smoothly and having the dashboard represent that was definitely something we had to look at twice in order to get right. Then the creative aspect where we had to design ourselves and see what type of information we want to show was also something we had to tackle early on.

Accomplishments that we're proud of

Being able to integrate many systems and have a planner scheduled for every member was really rewarding in the end. Seeing all the systems go in to play and having each AI complement each other was also something we were proud of. However, it wasn't only the back end we were proud of but all the design and graphs we were able to represent to give off a great user experience.

What we learned

We learned many libraries and tools throughout, especially the priceless experience we had by working with each other and being able to merge each others ideas and code. Also being able to understand the process that goes through funding and all the similar applications that an automatic caller could have. But also understanding the limits that some of our tools present and to not go beyond the scope that is realistically possible. However, now we have great understanding with how systems such as routing, react, motion, api keys, sqlite and other libraries go together to create Recoop.

What's next for Recoop

We’ll expand beyond flights and Amazon into hotels, rentals, subscriptions, and utilities, and add a clearer policy engine so eligibility and refund amounts are explainable. We’ll make price monitoring truly continuous with scheduled checks and alerts, plus add human-in-the-loop approval for higher-stakes calls. Finally, we’ll harden security with PII redaction, encryption, and audit logs, and ship a mobile experience so users can track savings anywhere.

Figma project link: https://www.figma.com/files/team/1597003740570051995/project/540272014/Team-project?fuid=1561412698763046283