Beacon

Beacon is an AI scribe for EMTs that captures live calls, distinguishes speakers, logs consent, and generates structured reports to reduce errors, improve handoffs, and support better patient care.

Comment 1

Keywords.ai Organizational ID

80571387-0305-416d-b692-c5fe06074818

X Link

https://x.com/TusharJain90235/status/2017858549674570028?s=20

Inspiration

As an EMT of 3 years, the feedback my coworkers and I receive on a quarterly basis has been surrounding the need for quicker contact to intervention times and faster documentation submission post-call. What administration doesn't realize is the deep value of moments spent empathizing with patient concerns, taking the time to properly think through every step of the interventions we provide, and holding their hand through what might be the worst day of their life. Beacon is a tool to accomplish just that, giving valuable time back to first responders and serving as a beacon of light in the broken US healthcare system.

What it does

Beacon is an AI-powered documentation assistant for first responders in EMS and firefighting contexts. It listens to live patient-provider interactions, differentiates between who is speaking, and only captures the voices of those consenting to the service with HIPAA-compliance capabilities assuming a transition to OpenAI's HIPAA-compliant model. Information voiced throughout the entirety of the call ranging from patient profile information to vitals obtained to interventions given are recorded and organized into the same categories used by existing EMS software such as ESO and Epic. However, this software goes over and above to generate patient condition trends observed in real-time, flag contraindications like potentially dangerous accidental drug interactions, and timestamps important time-sensitive interventions like IV medication administration and tourniquet application. Not only can EMTs engage in hands-free, high quality patient care, but they can use this tool to assist in informing their next steps and ensuring smooth handoff to the patient's next point of contact through the downloadable PDF feature.

How we built it

Beacon uses a multi-agent AI setup:

  • Listener Agent captures live conversations and differentiates between speakers
  • Medical Structuring Agent extracts vitals, symptoms, and interventions
  • Report Writer Agent generates structured JSON outputs that feed the dashboard and PDFs

Keywords AI orchestrates the agents and ensures reliability. GPT-4o-mini handles inference, Keywords AI monitors latency and Trae accelerates backend wiring, prompt iteration, and JSON parsing. Together, they enable fast, accurate documentation without slowing down crews.

Built With

Share this project:

Updates