Inspiration

270,000 Americans die from sepsis every year. Every hour of delayed treatment increases mortality by 8%. Early warning signs get buried in busy ICUs. We wanted to build an AI that catches what humans miss and acts immediately.

What it does

Sentinel monitors patient vitals in real-time, detects early sepsis patterns, and takes autonomous action - alerting care teams, pulling patient context, calculating risk scores, recommending treatment protocols, and documenting everything. The key feature: you can watch the AI think in real-time through a streaming reasoning panel that shows every step of its decision-making process.

How we built it

  • Yutori Research API to pull latest clinical guidelines and scrape patient data from hospital portals
  • TinyFish/Mino to auto-fill EHR documentation and extract structured data
  • Retool for the nurse command center dashboard
  • React + TypeScript frontend with streaming text animations
  • Claude API for clinical reasoning
  • Macroscope for code review during development

Challenges we ran into

Initially misunderstood what the sponsor tools did. Thought Yutori was for agent orchestration and TinyFish was for LLM reasoning. Had to pivot our architecture once we read the actual docs. Building the streaming text animation to feel smooth and premium took multiple iterations.

Accomplishments that we're proud of

The reasoning panel looks incredible. Watching the AI think through a clinical case in real-time with streaming text is genuinely compelling. The full detection-to-action loop runs in under 5 seconds. Integrated all 4 sponsor tools in ways that actually make sense for the use case.

What we learned

Read the docs first. Web automation tools (Yutori, Mino) are more powerful than expected for pulling data from systems without APIs. Building observability into the UI itself creates better user trust than hidden logging. Healthcare AI needs to show its work.

What's next for Sentinel

Real EHR integration via HL7 FHIR APIs. Expand beyond sepsis to cardiac events, respiratory failure, and other deterioration patterns. Mobile app for bedside alerts. Clinical validation studies. FDA clearance pathway exploration.

Built With

Share this project:

Updates