HeartGuardAI – Catch Heart Failure Before It Catches You

Video Call Between Doctor and Patient
Patients List With Statuses
Patients List

The Problem We Couldn't Ignore

My doctor friend told me recently that millions of heart failure patients silently slide into crisis, not because care isn’t available, but because they don’t realize they’re in danger.

Their bodies adjust. Symptoms creep in slowly. By the time they seek help, it’s too late, and the bill is $15,000+ per emergency hospitalization.

That’s not just broken -- it’s insane.

So we asked:

"What if a doctor could know a patient’s heart was failing… before the patient did?"

"What if we could see the danger before it happened, and act instantly?"

That’s why we built HeartGuard AI.

Inspiration

We started with a simple but powerful insight I once heard from my doctor friend:

“The body whispers before it screams.”

We saw that wearables and IoT devices already track data that patients ignore -- weight gain, shortness of breath, subtle biometric drift.

And we realized: AI could connect the dots faster than any human.

What it does

Our platform continuously evaluates patient data, detects decompensation patterns using AI models, and instantly triggers physician outreach.

Patients don’t have to do anything. The system watches. Doctors don’t have to guess -- they get objective alerts, not vague symptoms.

It’s like putting a guardian angel in the cloud, watching every heartbeat.

Doctor an patient can do a video call on the platform where AI listen to the conversation in realtime and provide suggestions to the doctor.

How we built it

Fronted – ReactJS
Backend – NodeJS/Express
Realtime communication – Websockets
Database – Sqllite
AWS IoT Core – for ingesting real-time health data from patient devices
Amazon Bedrock – to power our intelligent clinical decision-making layer
Video Call Integration (webrtc) – so a doctor can jump in the moment danger strikes
OpenAI Whisper – to give AI ability to "listen" what patient says and match with the metrics

What We Learned

"Real-time" matters -- catching a downward spiral even 1 hour earlier can change everything.
Trust = clarity + design -- we spent extra time on making the UI feel safe and intuitive.
AI isn’t replacing doctors -- it’s replacing delay.

The Hard Parts

Streaming and processing real-time IoT data without false alarms -- especially tricky for conditions with lots of noise.
Tuning LLMs to generate structured clinical alerts instead of vague suggestions.
Syncing AI decisions with real-world medical protocols in a way that physicians trust.

But the pain was worth it. Because...

The Outcome

HeartGuard AI isn’t just a hackathon project -- it’s a prototype for a future where no one dies because they didn’t know they were getting worse.

Each prevented hospitalization saves $15,000+ and weeks of suffering. 30% readmission rate within 30 days. Now multiply that by 6.5 million Americans with heart failure and $43B market.

We built this to win the hackathon. But the truth is -- if we get this right, everyone wins.