Inspiration
Chronic and rare disease patients don’t struggle because of a lack of data — they struggle because their data lives in silos, is hard to interpret, and is rarely actionable. Flare-ups often feel sudden, but in reality they usually build gradually. We wanted to build something that helps patients detect those subtle shifts early, understand why they’re happening, and walk into doctor visits prepared instead of overwhelmed.
CareConsole.ai was built around one simple idea: Give patients statistical clarity and clinical-grade summaries without making them feel like they’re doing research homework.
What it does
CareConsole.ai is a patient-centered health dashboard that helps detect symptom “flares” early.
Patients set a personalized baseline.
They log daily symptoms, sleep, and other lifestyle signals.
The system visualizes trends against that baseline with intuitive dashboards.
A client-side statistical engine runs:
Z-scores for deviation detection
EWMA smoothing to reduce noise
A weighted composite score to capture multi-symptom worsening
When sustained deviation is detected, it groups events into “flare windows” and explains exactly why the flare was flagged.
It also generates AI-powered clinician-style reports:
Structured medical summary
Plain-language explanation
Suggested questions to ask doctors
Reports are generated securely via backend API and are not stored.
How we built it
Frontend
React + TypeScript
Vite for fast dev iteration
Tailwind for rapid UI design
Recharts for trend visualization
Backend
Node + Express
MongoDB with Mongoose
JWT authentication with httpOnly cookies
Secure AI report generation endpoint
The flare detection logic runs entirely client-side to:
Reduce server costs
Preserve privacy
Improve responsiveness
We structured the app flow as:
Auth → Profile Setup → Baseline Onboarding → Dashboard (Overview, Daily Log, History, Insights, Reports)
The key architectural principle was separation of concerns:
Statistical logic on the client
Sensitive API keys on the server
No report persistence for privacy
Challenges we ran into
Noise vs real flare detection Symptoms fluctuate naturally. Early versions over-flagged. We refined the combination of Z-scores + EWMA + sustained deviation thresholds to reduce false positives.
Explaining statistical logic simply It’s easy to compute a composite score. It’s hard to explain it in plain language. We had to design “why this was flagged” outputs that feel transparent, not opaque.
Balancing privacy with functionality We wanted AI-generated reports but without storing sensitive health summaries. This required careful backend handling and stateless report generation.
Designing for cognitive load Chronic patients already deal with fatigue and brain fog. The UI had to be calm, clear, and non-overwhelming.
Accomplishments that we're proud of
A working, stable flare detection engine built in a short time frame.
Transparent statistical explanations instead of “black-box AI.”
A full end-to-end product flow from onboarding to AI-generated reports.
Client-side flare detection for privacy and scalability.
A clinician-style report that feels genuinely useful in appointments.
Most importantly, it feels like a product — not just a demo.
What we learned
Statistical signals matter more than fancy AI for early detection.
Transparency builds trust, especially in healthcare tools.
UI clarity is as important as model accuracy.
A composite score across symptoms is more reliable than single-metric tracking.
Privacy-conscious architecture decisions make the product stronger long term.
What's next for ConsoleCare.ai
Adaptive weighting based on user-specific flare history.
Multi-condition customization templates.
Smart baseline recalibration over time.
Clinician-facing export formats (FHIR compatibility).
Predictive modeling once enough longitudinal data exists.
Accessibility enhancements for low-vision and cognitive-fatigue users.
Mobile-first optimization.
Log in or sign up for Devpost to join the conversation.