MITRA

Inspiration

What it does

How we built it🌸 About the Project — Mitra

💡 Inspiration

Mitra was inspired by the harsh reality of domestic abuse in India, where nearly 30% of women face violence, yet less than 1% seek institutional help .

This gap is not due to lack of need—but due to:

Fear of legal complexity Social stigma around seeking help Lack of emotional and psychological support

We wanted to build something that doesn’t just provide solutions—but feels safe to approach. The idea was simple yet powerful:

What if help felt like talking to a trusted friend instead of entering a system?

This led to the creation of Mitra, an AI-powered companion that combines legal awareness, emotional support, and real-world action into one platform.

🛠️ How We Built It

Mitra is designed as a full-stack AI-powered web application with a strong focus on privacy, empathy, and accessibility.

🔹 Frontend Built using React.js + Vite for speed and responsiveness Styled with Tailwind CSS and Framer Motion for a calming UI Developed in TypeScript for better reliability 🔹 Backend Powered by FastAPI (Python) for fast and asynchronous APIs Uses SQLite databases for secure and lightweight storage 🔹 AI & Intelligence Layer RAG (Retrieval-Augmented Generation) for accurate legal guidance Groq API for fast LLM responses Sarvam AI for multilingual speech-to-text (Hindi, Hinglish, Nepali) 🔹 Core System Flow

The system works in a pipeline:

User shares their situation (text/voice) AI analyzes sentiment + context Legal rights are retrieved using RAG Response is simplified and delivered empathetically If risk is high → escalation to NGOs 🚀 What the Project Does

Mitra is not just a chatbot—it is a complete support ecosystem:

🤝 AI Companion: Provides emotional support like a trusted “Didi” ⚖️ Legal Guidance: Maps user situations to laws like IPC, BNS, PWDVA 🧠 Behavioral Detection: Tracks distress patterns using NLP 🏡 Anonymous Community: Safe peer-sharing space (Chautara Courtyard) 📊 Case Tracking: Seamless NGO integration and progress updates 📚 What We Learned

Building Mitra taught us that:

Technology Alone Isn’t Enough

We needed to combine:

Impact=Technology+Empathy+Accessibility

AI Must Be Responsible

In sensitive domains like law:

Hallucinated answers = dangerous Verified data (RAG) = essential

Language is Power

Supporting Hinglish and regional languages is not a feature—it’s a necessity.

Timing Matters

Survivors act within a “window of willingness”:

Help Probability∝ Delay in Response 1

Faster, non-judgmental responses significantly increase engagement.

⚔️ Challenges We Faced 🔹 1. Legal Accuracy vs Simplicity Legal texts are complex and intimidating We had to simplify without losing correctness 🔹 2. Avoiding AI Hallucination General LLMs can give wrong legal advice Solved using RAG with verified legal datasets 🔹 3. Sensitive UX Design UI had to feel: Safe Non-triggering Discreet 🔹 4. Behavioral Detection Identifying distress from text required: Sentiment tracking Pattern recognition over time 🔹 5. Privacy & Security Ensuring: Anonymous access No data misuse Safe usage in monitored environments