Saathi - Symptom awareness and Care Navigation tool

Saathi 

The Problem In India, menopause happens around 46.2 to 46.6 years of age—roughly five years earlier than the commonly cited global average. Yet, recognition without a route to care is incomplete. Specialist capacity is thin in many rural settings, meaning awareness alone is not enough. The first barrier to treatment is often not a medical one, but rather the ability to explain what is happening without shame. A woman rarely wakes up thinking she should complete a validated menopause questionnaire; instead, she uses everyday language, saying things like, "Aajkal raat ko bahut garmi lagti hai, neend nahi aati".

What is Saathi? Built by Team 404: Patriarchy Not Found, Saathi is a voice-first, India-first menopause-navigation companion. It is not a diagnosis app, a generic chatbot, or a simple tracker. Saathi is designed to turn short, natural language descriptions into a usable medical conversation starter. It meets women at their point of expression, transforming everyday Hindi and English into a clinically structured symptom summary and actionable next steps—without asking a woman to learn medical vocabulary first.

Key Features

• Voice-First Check-In: Users can complete a voice or text check-in using everyday language. No symptom checklist or clinical vocabulary is required to begin.

• Bounded Clinical Intelligence: The platform maps relevant phrases to the 11-item Menopause Rating Scale (MRS) symptoms and severity. It provides a structured picture but intentionally never calls this a diagnosis.

• Smart Uncertainty Protocol: If the system is unsure, it asks only low-confidence, relevant MRS questions. It presents a lower burden than an 11-question form and displays humility rather than AI overconfidence.

• Actionable Next Steps: Saathi delivers a plain-language summary, self-care education, and care routing. It effectively turns awareness into a usable next step.

• Doctor-Ready Summary & Family Bridge: The app generates a doctor-ready summary to improve the standard five-minute clinic visit. It also features an opt-in family card to help navigate household stigma.

How We Built It

• Frontend: Built using React, Vite, and Tailwind.

• Backend: Powered by FastAPI, featuring JWT auth, password hashing, and nearby doctor discovery via Google Places.

• AI/Voice Stack: Utilizes Browser Media Recorder/SpeechRecognition for voice input (with text fallback).

• NLP & Scoring: Features a Python NLP service that processes English, Devanagari, and Romanized Hindi against the 11 MRS symptom keys. It also uses a rule-based match combined with an optional, narrowly constrained LLM (Anthropic/Groq) to safely refine low-confidence keys.