We are not zombies

Pulzo: Bridging the Patient-Provider Gap

## Inspiration

Modern healthcare generates a staggering amount of unstructured data, from messy patient symptom logs to rapid-fire clinician voice notes. Much of this critical information remains "trapped" in text and speech, leading to clinician burnout and patients feeling disconnected from their own care plans. We were inspired to build a "Care Continuum" that turns this noise into structured, actionable medical intelligence.

## What it does

Pulzo is an end-to-end healthcare ecosystem that serves two primary users:

The Patient: Provides a digital health diary to log symptoms, mood, and food via voice or text. The AI provides immediate "wellness nudges" (e.g., "You've logged low sleep; consider resting") and visualizes health trends over time.
The Clinician: A high-speed voice-to-text pipeline that transforms raw medical dictation into clean, professional SOAP notes (Subjective, Objective, Assessment, Plan).
The Bridge: Our system automatically "translates" complex medical plans into simple, friendly language for the patient and summarizes the patient's weekly logs to pre-fill the "Subjective" portion of the doctor's clinical note.

## How we built it

We leveraged a robust .NET 8 backend with an asynchronous architecture to handle heavy AI processing:

Azure OpenAI & AI Foundry: Powering the core summarization logic and the complex mapping of raw text into the structured SOAP format.
Azure Speech-to-Text: Used for transcribing both casual patient logs and professional clinical dictations.
Text Analytics for Health: Crucial for Named Entity Recognition (NER) to automatically identify symptoms, medications, and dosages.
MongoDB (Azure Cosmos DB): Provided the flexible schema necessary for storing diverse health data and audit logs.

## Challenges we ran into

The primary challenge was the "translation" layer—ensuring the AI could accurately convert high-level medical jargon into empathetic patient advice without losing the clinical intent. We solved this through rigorous prompt engineering and utilizing Text Analytics for Health to bridge the terminology gap.

## Accomplishments that we're proud of

The "Closed Loop" Logic: We successfully built a system where a patient's daily life informs the doctor's visit, and the doctor's visit informs the patient's daily life.
Structured SOAP Output: Achieving high-quality, structured clinical notes from unstructured, rapid-fire voice input.
Clean Architecture: Implementing a C# backend that is scalable, auditable, and ready for real-world deployment.

## What we learned

We discovered that the "human element" in healthcare AI is just as important as the technical accuracy. Designing for "Nudges" taught us how AI can be used to support behavioral health, not just data entry.

## What's next for Pulzo

We plan to implement Azure AI Search to allow doctors to query years of patient history instantly and integrate with wearable device APIs to feed real-time vitals (like heart rate and sleep patterns) directly into the "Objective" section of the SOAP notes.