Skip to content
ATHARVA.SONARBack to portfolio

Project · AI · Safety Technology

SafeHer AI

Proactive personal-safety monitoring with risk scoring.

A full-stack safety platform that tracks live route context, scores risk with a scikit-learn engine, listens for spoken SOS phrases, and lets an agentic engine recommend or trigger protective actions as risk rises.

Live demoBuilt at Build And Grow: AI Hackathon 2.0
Live demo — SafeHer AI (opens in a new tab)Source code for SafeHer AI on GitHub (opens in a new tab)
SafeHer AI landing page reading 'Walk free. Stay protected.' with safety score, risk level and tracking status

What problem does SafeHer AI solve?

SafeHer AI helps people move with more confidence by combining live tracking, risk analysis, safe-route intelligence and emergency response in one app, so protective actions can start as risk rises instead of only after an SOS.

Key features

  • Live location tracking and safety dashboard
  • ML risk scoring (time, area, isolation, SOS)
  • Voice SOS keyword detection
  • Agentic responses by risk threshold
  • Safe-route and nearby safe-place lookup

How SafeHer AI works

  1. Start tracking. The user starts a tracked journey.
  2. Send context. The frontend sends location and context to the tracking API, which validates it.
  3. Score risk. The risk engine computes a score and the factors behind it.
  4. Respond by threshold. 50+ starts active monitoring, 70+ triggers high-risk recommendations, 85+ runs the full emergency workflow.
  5. Update the dashboard. The safety score (100 minus risk) and alerts refresh.

Architecture

  • Frontend: React and TypeScript (Vite, Tailwind CSS, shadcn/ui) dashboard with live route tracking.
  • API: Django REST Framework with Simple JWT, split into tracking, agents, alerts, routes and accounts apps.
  • Risk engine: A self-contained scikit-learn model scoring time of day, area type, isolation, location source, SOS status and safe-zone proximity.
  • Voice SOS: Keyword detection for phrases such as "help", "emergency", "bachao" and "madad".
  • Maps: Optional Google Maps APIs for safe routes and nearby hospitals, police and stations.
  • Data and hosting: SQLite or PostgreSQL; frontend on Vercel and backend on Render.

Tech stack

  • TypeScript
  • React
  • Vite
  • Django
  • Django REST Framework
  • scikit-learn
  • Google Maps
  • JWT

Engineering decisions

  • The ML risk engine needs no external model file, which keeps it easy to run and deploy.
  • When Google Maps is unavailable, the app falls back to default safe zones.
  • Production hardening (stricter access rules, host and CORS settings) is still in progress.

Atharva's role

Built with team Deploy House during Build And Grow: AI Hackathon 2.0.

Current status

Live demo at safeher-ai-five.vercel.app; core features are complete and production hardening is in progress.

Sources

Page updated by Atharva Sonar.

More projects by Atharva Sonar

  • SpendDecode — AI-powered personal finance and spending intelligence.
  • Velquix Learn — Structured roadmaps and curated resources for developers.
  • ClaimBlitz — Nine AI agents that process an insurance claim together.
  • FitHub Gym — Gym management with a computer-vision rep counter.
  • All projects