Inspiration
The inspiration behind SurgeSOS was born out of a critical need to address the alarming delays in emergency response times faced by callers. One such incident was for the residents of Oakland, California. A news article published in June 2023 shed light on the dire situation: when people dialed 911 in Oakland, they had to endure agonizingly long wait times of up to a minute before being able to communicate their emergency to a live dispatcher. These delays are not just frustrating but could potentially be life-threatening.
Alan Liang, an Oakland resident who experienced the ordeal firsthand, underscored the urgency of the situation: "I did call 911, and the experience was absolutely horrible. I was placed on hold for maybe about 10 to 15 minutes."
https://www.nbcbayarea.com/news/local/oakland-911-crisis/3266349/
What it does
SurgeSOS is a game-changing platform designed to help 911 centers when they get swamped with emergency calls. Our approach employs a multi-agent strategy: one agent engages with the caller, while another extracts vital information for dispatching units.
During times of high call volume, when traditional 911 lines become overwhelmed, callers would be transferred to SurgeSOS. We have ensured that communication is realtime so that caller can get quick responses.
Imagine a scenario where a caller, like Alan Liang, is in a dire situation, desperately seeking help. Instead of being placed on hold for minutes, they can now connect with SurgeSOS and have approiate units dispatched.
How we built it.
We had to carefully design our systems to make sure that essential tasks are carried out properly. We built a multi agent model where each agent specializes in its own task. We build this on top of open AI model and deployed it to Google Cloud Platform. We a;so build a flutter application to interact with the model. The flutter application synthesis speech and converts it to text and makes a POST request to the API. The API returns the response and the app converts it from text to speech. While the conversation is active, the agent is actively summarizing the conversation and extracting key information like location, situation and who to dispatch. This information is then stored in the Google Firebase Firestore Database and populated on our React website for easy reference and user experience.
Challenges we ran into
We ran into several challenges during this hackathon. We had to pivot from the original idea of using a contact number to voice call as there was a subscription cost. It was a challenge to get GPT 3.5 agents to behave like an actual 911 dispatcher, asking the right questions to the caller and extracting the right information from what they say. We also had challenges working with Firebase due the changes in API structure between different versions and database types. Finally, it was a challenge to connect the calling platform to the agents through Google Cloud Functions to remember the call history during the conversation.
Accomplishments that we're proud of
We are proud to have successfully built and validated a proof-of-concept AI emergency response agent that promises similar experiences to a 911 emergency call.
Moreover, We are proud to have built an end-to-end solution that improves a real-world emergency response system. We believe that our solution can help reduce high caller wait times and provide some support to our 911 operator heroes. Our multi-agent model performed well, and we successfully integrated multiple platforms into a unified architecture. This is also our first time working with Voice, and we’re quite happy with the results.
What we learned
We have never worked with voice synthesis and speech to text integrations until this project, and we’re happy to have to working properly. It was also the first time for a couple of us to make use of React and connect a backend like Firebase to it.
This project gave us a new appreciation for all the work 911 dispatchers do. It takes so much knowledge, skill and compassion to handle back-to-back emergency calls all day. While we're proud of SurgeSOS, humans will always be irreplaceable when lives are on the line.
What's next for SurgeSOS
We are excited by SurgeSOS's early promise, we see this technology to be transformive during high traffic times. We hope our solution sparks new ideas and collaborations between the tech community and public safety professionals.
Although SurgeSOS may not be deployed exactly as is, the possibilities are boundless for how it can be improved upon. Our ultimate vision is ensuring anyone in urgent need can swiftly get skilled emergency assistance, no matter the time or place.We are optimistic about a future where no 911 call goes unanswered, and all people feel equally protected. SurgeSOS represents an important step toward that ideal. But there is so much more innovative potential to unleash.
We were deeply moved by stories of individuals in dire emergencies being placed on hold for precious minutes when they called 911. When lives are at stake every second counts.
Log in or sign up for Devpost to join the conversation.