Inspiration
I was first inspired to build 911 AI through witnessing a friends experience with calling 911 after an accident. While the accident was minor the time delay would have had serious consequences if luck had not been on his side. However as I learned more it became obvious to me that the current 911 system was in need of improvement but was slow to change. 911 Responder is an attempt to take technology and build a system that supports rather than replaces our current 911 system, in the process creating in impact where none had existed before.
What it does
911 Responder is a web based tool focused on improving the efficiency of 911 dispatchers. From anywhere in the world a user can call our servers where an AI system will greet them and ask them questions related to their current emergency. Through doing so it will cycle through several questions before extracting the most important details of that emergency. This information will then be put onto the user interface allowing the 911 Dispatcher to see specific details of the emergency that can then be used in whatever course of action the dispatcher decides on or in reaching out to the user again.
It is important to note that throughout the process the 911 Responder will only be active in scenarios where no human dispatcher will be available. In addition the dispatcher will be able to interject, decide, and respond to any call as they see fit with this build serving to save the time that would be lost if no dispatcher was available and the caller was forced to wait in silence.
How we built it
This build was constructed largely with Node.js in addition to Twilio which allowed the build to answer calls and verbally answer questions by the user. As to these responses ChatGBT and its APIs are used in generating responses based on prompt engineering with various prompts informaing ChatGBT what role it should play and what position in the conversation they are in. Finally Ngrok is used to facilitate interaction between the website, private backend, and the public Twilio accounts allowing for seamlessly integration between functionalities running on local hosts and the wider internet. Finally the hugging Face API is used to build several ancillary functions such as All-MiniLM-16-v2 for basic sorting.
Challenges we ran into
When building 911 Responder the main challenge I faced was getting the various API services to work together. This was largely due to the extreme importance of the backend in my build as large portions of the frontend relied on information provided with many API calls. During my hacking I was forced to change and quickly learn several new APIs in order to finish the project with the greatest challenge being the integration of Ngrok. Originally this API was not part of my plan however as the back-end began to become more realized I had to take the effort of quickly learnings its applications, definitely one of the key experiences of the hackathon for me.
Accomplishments that I'm proud of
One crucial accomplishment I am proud of has to do with the integration of the front and backends. During the process of building 911 Responder I was forced to change my plans several times with the greatest difficulty being wether I could integrate the front and backend into each other. With every error I found myself doubting my ability to execute on the project however through quick learning I was able to utilize most of my desired features.
What I learned
The most important lesson I learned through the process of building this product is the value of a development schedule especially for larger projects with several moving parts. For example the process of learning several APIs resulted in me needing to understand the impact of that process on the completion of my project. Knowing how much time and energy I would devote to each feature would then allow much faster decision making as to what features I should implement and how to do so. This lesson was then further reinforced through my conversations with mentors who stressed the importance of said planning in their own professional organizations, making this a skill I highly value.
What's next for 911 Responder
Next up for 911 Responder I will be finalizing the secondary and tertiary applications of 911 Responder being its 311 and Business call center applications. As explained in my pitch this will allow me to develop this idea to a greater extent by providing immediate use cases and resources for development. In addition I will need to expand the use of the ChatGBT Whisper and Hugging Face API as the former was taught to be by a mentor during the conference while the latter supports several features that could be expanded.
Log in or sign up for Devpost to join the conversation.