SympSense: Safe & Observable Health AI

A transparency-first AI health assistant that transforms symptom checking into structured, explainable, and calm decision support.

Comment

🌌 Inspiration

During exam seasons and high-stress academic periods, students often experience physical discomfort but struggle to understand whether their symptoms are temporary, stress-related, or require medical attention.

Existing health tools typically fall into two extremes:

  • Overly complex platforms filled with medical jargon
  • Generic AI chatbots that provide answers without clarity or accountability

We initially set out to build a calm, student-first symptom awareness tool.

However, during development, we identified a deeper issue:

Most AI systems in healthcare operate as black boxes — users receive outputs, but have no visibility into how or why those outputs are generated.

This led us to rethink the system itself.

👉 SympSense evolved into a transparency-first health AI assistant focused on clarity, safety, and trust.

🚀 What it does

SympSense is a non-diagnostic AI health assistant that helps users:

  • 🧠 Describe symptoms naturally using text or voice
  • 💬 Receive clear, structured, easy-to-understand explanations
  • ⚠️ Understand whether symptoms can be monitored or require attention
  • 🚨 Get guidance for potentially serious conditions
  • 🌍 Access multilingual support for improved accessibility

🔥 What makes SympSense different

Most AI health tools optimize for generating answers.
SympSense is designed to optimize for understanding, safety, and user trust.

🧠 Structured AI Understanding

Instead of long, unstructured responses, outputs are organized into:

  • Clear explanation
  • Severity level (Low / Moderate / High)
  • Confidence indicator
  • Recommended next steps

👉 This reduces cognitive overload and helps users make clearer decisions.

🕒 Symptom Timeline & Pattern Awareness

  • Tracks symptoms over time
  • Identifies recurring patterns
  • Enables continuous health awareness

👉 Moves beyond one-time queries to longitudinal understanding

📊 Observability Layer (Core System Thinking)

SympSense introduces an observability layer that monitors:

  • Response latency
  • System behavior
  • Safety-related signals

👉 This shifts the AI from a black-box responder → a monitored system

This serves as a foundation for deeper reliability and safety evaluation.

🛡️ Safety & Transparency Layer

  • Detection of potentially high-risk symptoms
  • Escalation guidance for critical scenarios
  • “Why this response?” explanation panel
  • Clear non-diagnostic boundaries

👉 Designed to inform responsibly without creating panic or misinformation

🧘 Calm, Human-Centered UX

  • Minimal, soft interface designed to reduce anxiety
  • Emotion-aware tone and microcopy
  • Accessible and inclusive interaction design

👉 Focused on supporting users during uncertainty

⚠️ Important: SympSense does not provide medical diagnoses. It supports awareness and encourages consultation with healthcare professionals.

🏗️ How we built it

Frontend

  • TypeScript, HTML, Tailwind CSS
  • Component-based architecture
  • Mobile-first, responsive design

AI & Backend

  • Google Gemini API for natural language understanding
  • Lightweight backend for secure AI interaction
  • Structured response formatting layer

Observability System

  • Datadog integration for:
    • Request/response logging
    • Latency tracking
    • System behavior monitoring

🔄 Finale Round Improvements

To align with the refinement-focused finale round, we prioritized depth and system design:

  • 🔄 Refactored into a more scalable architecture
  • 🧠 Introduced structured AI outputs (severity, confidence, actions)
  • 🕒 Added symptom timeline and pattern tracking
  • 📊 Integrated observability monitoring
  • 🚨 Strengthened safety and escalation handling
  • 🎨 Improved UI/UX for clarity and calm interaction
  • 🌍 Enhanced multilingual accessibility

👉 The system evolved from a basic chatbot into a more structured and responsible AI experience

⚙️ Challenges we ran into

  • Designing AI responses that inform without increasing anxiety
  • Maintaining non-diagnostic boundaries while remaining useful
  • Defining meaningful observability beyond simple logging
  • Balancing system monitoring with user privacy
  • Avoiding over-engineering while preserving usability

🏆 Accomplishments that we're proud of

  • Built a production-style AI system, not just a prototype
  • Introduced structured outputs for better user understanding
  • Integrated observability into an LLM workflow
  • Designed a calm and accessible experience for sensitive use cases
  • Balanced intelligence, safety, and usability effectively

📚 What we learned

  • AI systems in healthcare must be understandable, not just accurate
  • User trust depends on both UX design and system transparency
  • Responsible AI requires system-level thinking, not just model integration
  • Observability enables proactive trust, not just debugging

🔮 What’s next for SympSense

  • Advanced confidence calibration and reliability scoring
  • Hallucination awareness and response validation
  • Long-term personalized health insights
  • Clinician-facing dashboards for oversight

Real World Impact

SympSense addresses a global challenge in digital healthcare:

• Lack of trust in AI-driven medical tools
• Over-reliance on black-box systems
• Anxiety caused by unclear or overwhelming outputs

By introducing explainability, structured outputs, and observability, SympSense improves:

• User understanding and confidence
• Responsible AI adoption in healthcare
• Early awareness and better decision-making

This makes it not just a tool, but a foundation for safer AI systems in real-world healthcare environments.

Built With

Share this project:

Updates