DisasterDesk

Inspiration

After doing a bit of research on the topic, we found that over 80% of call centers are understaffed and overworked, especially during times of crisis. During crisis, there are calls about every 7.6 seconds, causing wait times to skyrocket up to multiple hours, leaving those in life or death scenarios on hold, waiting for someone to respond to them.

Many of these calls that go in are duplicate and most times non life-threatning; Clogging up the phone lines for actual people who need immidiate help.

Currently, what these call centers do to try and amend this issue is to use natural language processing to try and analyze what callers are saying. However this is quite unreliable and is not a viable solution during times of crisis as it does little to open the bottleneck of calls.

What it does

DisasterDesk solves this problem by using an agentic ai that will automatically interpret calls and categorize them into different severity ratings, letting the calls of people who are in life or death scenarios go through and condesing the multiple calls reporting the same thing down to a single message that can be sent to 911.

How we built it

We built this using Nextjs, React, TailwindCSS, Prisma, and Google Gemini.

Challenges we ran into

Text to Speech delay: When using Google Gemini to do the text to speech, the response times were too long and made it feel unnatural and clunky, which could cause more stress to someone who is in a crisis. To solve this problem, we used a more basic text to speech which sped up times greatly.

Location Grabbing issues: When we prompted the ai to talk to the user, it would try to guess the users location, giving an estimate, which was often incorrect. To fix this we used a more precise location api that triangulates the users location.

Accomplishments that we're proud of

Our clean front end that makes it simple and easy to use. We are also proud of coming up with a solution that could help people nationwide and is simple, yet effective at addressing the problems that show up during times of crisis.

What we learned

We learned how to use Google Gemini and how to get location data.

What's next for DisasterDesk

Connecting it up to local 911 call centers for testing and refining the product. We also want to expand on the scope of the project by integrating with other relief organizations and support systems.