Cardio Command

Inspiration

Cardiac recovery does not end at discharge. The most fragile part often begins when the patient goes home, away from continuous clinical supervision. During that period, patients are managing pain, medication schedules, activity limits, anxiety, and early warning signs, while care teams have limited visibility into what is happening day to day.

We built Cardio Command to close that gap. Our idea was to create a connected recovery system that supports both sides of care: a clinician command center for monitoring and decision support, and a patient experience that feels simple, motivating, and available at home. The latest version pushes that even further by bringing in rehab coaching and wearable sync, so recovery is not just monitored, but actively guided.

What it does

Cardio Command is an AI-powered post-cardiac recovery platform.

It includes:

A clinician dashboard with live vitals, AI risk analysis, alerts, and patient-level monitoring
A patient app that translates recovery data into simple, supportive language
AI chat and voice support so patients can report symptoms and get guided next steps
Recovery plan and rehab coaching flows that turn doctor instructions into patient-friendly support
Rehab streak tracking and progress feedback to encourage adherence
WHOOP integration to bring in recovery, sleep, strain, resting heart rate, HRV, and workout context
A shared backend that syncs patient updates, alerts, and analysis in real time

The result is a more complete recovery experience: not just passive monitoring, but active support between surgery and stability.

How we built it

We built Cardio Command as a full-stack monorepo with two frontends and one shared backend.

Frontend

We created two React apps:

A clinician-facing dashboard
A patient-facing mobile-first app

We used:

React
Vite
JavaScript
Tailwind CSS
Framer Motion
Recharts

On the patient side, we recently added a stronger recovery engagement layer with rehab streaks, calendar-style progress, celebratory feedback, and wearable connection states. On the clinician side, we added workflows for writing recovery plans that can be transformed into AI-guided patient coaching.

Backend

Our backend is built with FastAPI and handles:

Patient and vitals APIs
Real-time WebSocket updates
AI endpoints
Voice workflows
Recovery plan persistence
WHOOP OAuth and sync endpoints

We use SQLAlchemy with SQLite for persistence, and we added models for call history, recovery plans, and WHOOP connections so the app can maintain longitudinal recovery context.

AI and data layer

We combined several systems to make the AI more useful and grounded:

LangGraph for orchestrating multi-step analysis
FAISS for retrieval over clinical guidance documents
OpenAI GPT-4o for summaries, coaching, chat, and documentation
A deterministic risk model for explainable scoring
Rehab state tracking to record wins, barriers, and progress over time

This helped us build an experience where the AI is not just reacting to symptoms, but also supporting adherence and recovery behavior.

Challenges we ran into

One challenge was balancing two very different user experiences.

Clinicians need a dense, fast interface with high-signal information. Patients need a calm, simple, encouraging experience. That became even more important once we added rehab coaching and wearable sync, because recovery data can easily become overwhelming if presented the wrong way.

Another challenge was making the system feel connected in real time. Vitals, AI outputs, voice workflows, recovery plans, rehab progress, and WHOOP data all had to feel like parts of the same product rather than disconnected features.

We also had to think carefully about trust. In healthcare, we wanted to avoid a pure black-box experience, so we paired LLM outputs with deterministic logic, retrieval-based grounding, and structured patient state.

Accomplishments that we're proud of

Built a true two-sided recovery platform for both clinicians and patients
Added real-time monitoring and AI-supported clinical workflows
Expanded the patient experience beyond symptom reporting into rehab motivation and recovery adherence
Integrated WHOOP wearable sync with OAuth and recovery data summaries
Added doctor-authored recovery plans that can power more personalized patient coaching
Created a demo that shows a full loop from patient data to clinician action to patient guidance

The newest updates made the project feel much more like a real recovery companion, not just a dashboard.

What we learned

We learned that healthcare AI becomes much stronger when it supports behavior, not just analysis.

A patient does not only need to know that something is wrong. They also need help staying on track, understanding their data, and feeling supported through recovery. Adding rehab tracking and wearable integration reinforced that idea for us.

We also learned:

Real-time visibility is powerful, but motivation and follow-through matter just as much
Wearable data is most useful when translated into context, not just numbers
Patient-facing AI needs a very different tone from clinician-facing AI
Explainability is essential in healthcare products
Small product details like streaks, progress, and coaching can meaningfully improve engagement

What's next for Cardio Command

Next, we want to take Cardio Command from a strong prototype to a more production-ready recovery platform.

Our next priorities are:

Expand wearable integrations beyond WHOOP
Personalize rehab coaching based on recovery phase, symptoms, and clinician plans
Improve escalation logic for urgent symptoms and high-risk patterns
Add stronger clinician communication and follow-up tools
Strengthen compliance, privacy, and auditability for real-world use
Pilot the experience around actual post-discharge cardiac recovery workflows

Long term, we want Cardio Command to become a continuous recovery layer that helps patients feel supported at home and helps care teams intervene earlier when it matters most.

                Cardio Command

    +------------------+      +------------------+
    |   Patient App    |      |  MD Dashboard    |
    |------------------|      |------------------|
    | - Vitals view    |      | - Live monitoring|
    | - Rehab streaks  |      | - Risk alerts    |
    | - AI chat/voice  |      | - AI analysis    |
    | - WHOOP sync     |      | - Recovery plans |
    +--------+---------+      +---------+--------+
             \                         /
              \                       /
               \                     /
                +-------------------+
                |   FastAPI Backend |
                |-------------------|
                | - APIs            |
                | - WebSockets      |
                | - Voice workflows |
                | - Rehab tracking  |
                | - WHOOP OAuth     |
                +---------+---------+
                          |
                          |
      +-------------------+-------------------+
      |                   |                   |
      v                   v                   v