The Problem

Every year, millions of people turn to LLMs for health advice. While these models are great at providing general information, they face a critical "last mile" failure: they are disconnected from the actual healthcare system.

Patients are left with a fragmented experience:

The "Advice Trap": While 80% of internet users search for health information online, LLMs are limited to general advice. They can't help you get an appointment with an actual doctor when it is needed.
Manual Burden: Patients are forced to navigate complex phone trees and wait on hold, a process that can take 10–20 minutes per call just to check availability.
Information Silos: Despite the availability of AI, those models aren't linked to healthcare systems, leaving users to manage their own triage and logistics.

There is a desperate need for a personal health assistant that doesn't just talk, but acts.

The Solution

MedRedemption is an autonomous personal health assistant that bridges the gap between AI advice and clinical action. It doesn't just tell you to see a doctor; it finds the clinic, makes the phone call, and secures your spot—all through a seamless conversation.

How it Works

Multimodal Diagnosis: Users engage in a natural, voice-based dialogue with MedRedemption, which can also "see" physical symptoms via camera (powered by Gemini 2.5 Flash).
Intelligent Triage: A dedicated reasoning agent (Llama 3.3 on Groq) analyzes the conversation in real-time, calculating a medical risk score (0-600) to detect urgency and the specific need for an in-person visit.
Autonomous Triggering: When the system detects a "Book Routine" (Code 400+) or "Emergency" (Code 600) status, it automatically transitions from advice to action.
Clinic Discovery: MedRedemption uses OpenStreetMap and Tavily to geolocate the user and find the nearest open walk-in clinics and their contact numbers.
Agentic Telephony: MedRedemption initiates a real-time voice call to the clinic using ElevenLabs Conversational AI. It speaks to the receptionist, negotiates an appointment time, and confirms availability.
Closing the Loop: The assistant reports back to the user with the secured options to finalize the booking.

How We Built It

We built MedRedemption as a sophisticated multi-agent system:

The Intelligence Layer

Google Gemini 2.5 Flash (Multimodal Live API): handles the primary user interaction, processing real-time audio and video frames to understand symptoms and maintain a natural conversation.
Groq (Llama 3.3): Serves as the high-speed "decision engine," monitoring the chat for medical urgency signals and strictly adhering to safety protocols (e.g., advising emergency care for high-risk scores).

The Action Layer

LangChain & Tavily: Orchestrates the search for real-world healthcare providers, utilizing IP geolocation and OpenStreetMap data to find verified clinics near the user.
ElevenLabs Conversational AI: Powers the telephony agent that physically calls clinics. It uses low-latency text-to-speech and speech-to-text to have human-like conversations with receptionists.

The Infrastructure

Python (Flask) & Go: A hybrid backend where Python handles the agentic logic and a high-performance Go server manages the telephony streams.
Next.js: Provides the user-facing dashboard for managing health profiles and viewing appointment history.

Accomplishments We're Proud Of

Real-World Agency: We broke the "chatbox barrier" by enabling an LLM to make actual phone calls and navigate real-world booking flows.
Safety-First Architecture: Implemented a rigorous risk scoring system (0-600) that ensures the AI escalates emergencies appropriately rather than just hallucinating advice.
Multimodal Empathy: By using Gemini 2.5's vision capabilities, MedRedemption can look at a user's injury (via camera) and offer more grounded, empathetic responses than text alone.

What’s Next for MedRedemption

Direct EHR Integration: Securely syncing with electronic health records to provide clinics with pre-visit summaries.
Insurance Verification: Automatically checking if a clinic accepts the user's specific insurance provider during the call.
Follow-up Autonomy: Automatically calling the patient 24 hours after their appointment to check on their recovery.

MedRedemption: Because health advice is only helpful if it leads to healthcare.