Medi Paru - Medical Concierge Agent

Landing Page
Reception Main Feature
Data session details
Medical record information synthesis and analysis
QR record
Via QR to be able to export medical records everywhere

Inspiration

Healthcare systems, especially in busy hospitals or clinics, often face long waiting times and inefficiencies in the initial stages of patient intake. Many patients spend unnecessary time explaining basic symptoms and history, which can be automated. Inspired by the need to reduce patient wait time and streamline pre-consultation procedures, we created Medi Paru, a virtual medical concierge agent that serves as the first point of interaction between patients and healthcare providers.

What it does

Medi Paru is a conversational AI agent designed to simulate a medical office assistant. It interacts with patients, gathers essential health information, and summarizes the case into a structured medical note for doctors. Key functionalities include:

Greeting and collecting basic patient info (name, year of birth, gender)
Checking for returning patients and reviewing past medical records
Sequentially asking medically relevant questions across various categories
Extracting symptoms and health history without providing medical advice
Handling document uploads (e.g., prescriptions, lab results, discharge papers)
Generating a structured medical summary ready for physician review
Suggesting an appropriate medical specialty based on gathered data

How we built it

We built Medi Paru using:

Large Language Model (LLM) prompting techniques for empathetic and context-aware conversations
A custom state management system to track medical consultation stages
Function calling capabilities to handle logic like saving medical notes, checking patient history, and finalizing records
File uploader integrations to collect medical documents for more accurate diagnosis
Guardrails and ethical constraints to prevent the agent from giving medical advice, ensuring it's only a data collection assistant

Challenges we ran into

Designing a natural conversation flow that remains medically structured but human-friendly
Ensuring data integrity and accuracy when users provide incomplete or ambiguous information
Creating a system that is empathetic but non-diagnostic, respecting the ethical boundaries of AI in healthcare
Handling edge cases where patients don't cooperate or skip important questions
Managing state tracking over a potentially long conversation without losing context

Accomplishments that we're proud of

Built a complete end-to-end virtual consultation workflow without needing medical professionals to intervene at the intake stage
Successfully integrated contextual prompting to collect complete health data across multiple categories
Designed a fail-safe system where the agent gracefully handles refusals or incomplete answers
Improved time efficiency for doctors by providing them with a ready-to-review medical summary
Made the interaction empathetic and user-friendly, even for non-tech-savvy patients

What we learned

The importance of conversation design in sensitive domains like healthcare
That users can behave unpredictably, and systems must be designed to handle graceful fallbacks
Prompt engineering can drastically change the tone, performance, and compliance of AI agents
Function calling and context updates are critical in turning a conversational agent into a structured assistant
The balance between automation and ethical constraints is key in medical applications

What's next for Medi Paru - Medical Concierge Agent

Integrate with Electronic Health Record systems for seamless data flow into hospital systems
Implement voice interaction and accessibility features
Conduct pilot studies with clinics or hospitals to measure real-world impact
Improve with real-time learning feedback loops, enhancing accuracy and experience over time