PulseNet

Inspiration

In an emergency, reaching the nearest hospital does not always mean reaching the right hospital. A patient's condition can deteriorate during transport, while ICU beds, ventilators, specialists, and other resources can change continuously. We were inspired by this gap in emergency coordination and wanted to build a system that considers both the patient's current condition and the hospital's current readiness, rather than relying only on distance.

What it does

PulseNet is an AI-assisted emergency decision-support platform that continuously evaluates patient condition and hospital readiness to recommend the most suitable receiving hospital and required resources in real time.

For example, if a patient has chest pain and falling SpO₂, PulseNet can compare nearby hospitals based on factors such as ICU availability, specialist availability, required equipment, and estimated arrival time. The recommendation can then adapt when the patient's condition, hospital resources, or ETA changes.

How we built it

We designed PulseNet around a real-time emergency coordination workflow. The system consists of responsive dashboards for paramedics and hospitals, a backend using REST APIs and real-time event processing, and a real-time communication layer using WebSockets/socket-based communication.

The decision engine uses explainable emergency triage rules for the MVP, with ML-based deterioration-risk prediction planned as the intelligence layer. Patient state, hospital capacity, resource availability, and live ambulance telemetry are used to support resource-aware hospital ranking.

For the initial prototype, the workflow can be demonstrated using simulated patient vitals, ambulance locations, hospital capacity, ICU/ventilator availability, and specialist availability.

Challenges we ran into

One of the biggest challenges was designing a system where a hospital recommendation is not based on a single factor such as distance. Emergency decisions involve multiple changing variables, including patient severity, hospital resources, specialist availability, and ETA.

Another challenge was making the system real-time while keeping its decisions explainable. We wanted the recommendation to be understandable to emergency teams rather than functioning as a black box. We therefore focused the MVP on explainable rules and human-supervised decision support.

Accomplishments that we're proud of

We are proud of creating a concept that moves emergency coordination from reactive searching to proactive preparation.

PulseNet brings patient condition, ambulance information, and receiving-hospital readiness together in one workflow. It can provide dynamic destination guidance while enabling hospitals to prepare for an incoming patient with information such as condition, ETA, required resources, and urgency level.

What we learned

We learned that solving a healthcare problem is not only about building an AI model. The system must fit into a real-world workflow, provide timely information, remain understandable to its users, and account for constantly changing conditions.

We also learned the importance of starting with a feasible MVP. By using simulated data first, we can validate the complete workflow before progressively integrating real-world clinical and hospital systems.

What's next for PulseNet

Our next step is to turn the prototype into a working end-to-end system with live patient and ambulance simulation, real-time hospital resource updates, and dynamic hospital ranking.

We also plan to develop the ML-based deterioration-risk prediction layer and progressively integrate real-world clinical and hospital systems. The long-term goal is to make PulseNet a reliable, explainable, and human-supervised platform for improving emergency-care coordination.