ZeroResponders

Inspiration

In rural and developing nations, people can often struggle to get medical attention in time. Extremely preventable accidents, such as heart attacks, can be prevented with something as simple as CPR. There is a need for a way to connect volunteers with people in need of medical attention.

What it does

ZeroResponders provides a platform for people to share medical data with EMS and for volunteers to respond to emergencies they feel comfortable dealing with. On-going emergencies are displayed on a convenient Google Map, allowing people to get instant directions to assist the victim.

When a user reports an emergency for themselves, ZeroResponder will automatically contact the police with the user's medical information as well as location. The user's medical data is run through a neural network to detect heart disease. This can provide extreme accuracy for the first responders who need to aid the patient. If the user becomes incapacitated, the dispatcher can call back and talk to a GPT-3 assistant given the emergency data.

When a user reports an emergency for someone else, ZeroResponder requests some basic medical information that someone like their family member would typically know, however, it is entirely optional. It then goes through the same process of dispatching EMS, alerting nearby volunteers, and letting police speak to a GPT-3 assistant in case the bystander is too flustered or must attend to the victim.

How we built it

ZeroResponder is built in React Native with a Firebase BaaS and neural networks trained in Python. The server that manages calls to the First Responders is built on Twilio.

Challenges we ran into

Overfitting was a huge problem when training the neural network, to get around this we added Dropout layers which randomly dropout variables to avoid developing biases.