Caddie

Inspiration

Emergency call center operators are always confronted with calls that are hard to understand because of a thick, unclear accent, loud background noises, a panicked or rapid tone, or other factors.

What it does

Caddie is an answer to these problematic situations, where every second counts and lives are possibly at stake. The operator either inputs the call's audio file or records it in real time. Caddie transcribes this audio and processes it to give the operator a summary of the call, a severity rating, and actions that can be recommended to the caller. This allows operators to waste less time trying to understand the situation, and more time trying to solve it.

How we built it

We built Caddie using React, Vite, and TailwindCSS for the front end. Caddie also uses Cloudflare's Workers AI service to transcribe the inputted audio (utilizing the full power of a Whisper-based solution, which is very effective at eliminating background noise) and process it to generate the desired output for the user.

Challenges we ran into

As we had never worked with Cloudflare before, figuring out how to correctly utilize its AI services required a lot of time and effort.

Accomplishments that we're proud of

We are very proud of successfully implementing AI audio transcribing and text processing/generating technology into our solution.

What we learned

We learned that while using AI in a project can be daunting and may take some time to figure out, it is far from being impossible. We are now more confident in our abilities to correctly use the power of AI to better the world around us.

What's next for Caddie

The immediate next step for Caddie would be to implement a feature enabling the user (operator) to receive, transcribe, and process multiple calls at once. This would utilize the full potential of ranking calls using their assigned severity ratings which represents the urgency of the situation.