AI-Enabled Wearable Safety · An Initiative of G.I.R.L.S. Organization

HAVENKEY

HAVENKEY

Safety that doesn't wait for you to act.
Passive distress detection: before danger becomes obvious.

See the Solution Try the Live Demo G.I.R.L.S. Community
1 in 3
Women experience violence in their lifetime (WHO 2025)
$120B
Wearable Safety TAM, growing 12% CAGR through 2030
Top 3%
Blue Ocean Competition · 12,784 global submissions
30+ States
G.I.R.L.S. pilot network across 17 countries

The Problem

Safety tools require you to act. That's the flaw.

Most safety apps depend on a girl deciding the danger is bad enough to escalate, then unlocking her phone, opening an app, and pressing a button. In the moments that matter most, that sequence is impossible. Existing wearables are reactive, conspicuous, and dependent on user initiation, which fails exactly when protection is needed most.


Through the G.I.R.L.S. Organization, I have heard members describe abandoning internships, competitions, and travel opportunities because of safety risks their male peers never had to weigh. The problem isn't lack of ambition. It's a gap in protection.

640M
Women worldwide live in fear inside their own homes
37%
IPV prevalence in South Asia, the highest globally
<50%
Of survivors seek help due to stigma and isolation
0 sec
Available to act when danger is immediate

Our Solution

Passive. Intelligent. Automatic.

HavenKey is an AI-enabled wearable that performs continuous distress detection through multimodal sensor fusion, analyzing voice biomarkers, motion and accelerometer data, and physiological signals such as heart rate variability, without requiring a panic button or an unlocked phone.


The core classification model correlates signals across all three inputs simultaneously, because danger rarely shows up in just one signal alone.

FeatureExisting ToolsHavenKey
ActivationManual panic buttonAI-automatic detection
Phone requiredMust unlock deviceIndependent GSM/GPS
Visibility to threatConspicuous useDiscreet wearable form
False positivesSingle-threshold triggersAdversarially tested ML
Post-crisisOne-time alertContinuous monitoring loop
🎙

Voice Biomarker Analysis

Detects distress in ambient audio without requiring the user to speak or call for help.

📡

Motion & Accelerometer Fusion

Reads chaotic motion, sudden impacts, and prolonged stillness, which are signatures of real danger.

❤️

Heart Rate Variability

Monitors physiological stress response to distinguish fear from exercise or daily activity.

⚡

Real-Time Alert Dispatch

Autonomously notifies trusted contacts and emergency services within seconds of confirmed distress.

📶

Bluetooth & GSM Connectivity

Operates via Wi-Fi, Bluetooth, or GSM and fully functional independent of the user's phone.

🔋

48-Hour Battery Life

Continuous protection during extended emergencies, travel, or low-signal environments.

How It Works

Three inputs. One label. No action required.

The classification model correlates voice, motion, and heart rate simultaneously, because danger rarely shows up in just one signal.

01

Sensor Data Collection

HavenKey continuously reads accelerometer, gyroscope, pressure, GPS, and ambient audio, passively, with no user interaction required.

02

ML Pattern Recognition

A Python-based classification pipeline fuses sensor data across time windows, stress-tested against adversarial false-positive scenarios including erratic motion, elevated heart rate from exercise, and loud ambient environments.

03

Risk Evaluation & Alert

When confirmed distress is detected, HavenKey dispatches a precise GPS alert with distress coordinates to trusted contacts and emergency services within seconds, with a subtle vibration confirming dispatch.

Detection Architecture

A continuous loop, not a one-time trigger.

HavenKey doesn't fire once and stop. The feedback loop runs continuously, updating risk scores and re-evaluating context in real time.

1

Sensor Data Collection

Continuous passive reading across all input channels

2

Real-Time Data Processing

Feature extraction across short and long sliding windows

3

Pattern Recognition

ML model correlates multimodal signals into a joint risk score

4

Risk Evaluation

Pathway-branched confirmation rules differentiate distress types

5

Emergency Mode Activation

Alert state machine escalates through SUSPICIOUS → ALERTING

6

Priority Emergency Comms

GPS coordinates dispatched to trusted contacts and services

7

Continuous Live Tracking

Location updates continue until manual cancel or cooldown

8

Post-Incident Support

Session logged for review; model refines from labeled outcomes

Built under G.I.R.L.S. Organization

HavenKey emerged from four years of running a global STEM movement. The G.I.R.L.S. Organization, founded in in 2022, now active across 30+ U.S. states and 17 countries, is both the reason HavenKey exists and its first pilot community. Members will wear, test, and help label the training data that shapes the model's next iteration.

Visit G.I.R.L.S. →

Recognitions & Affiliations

Validated by peers, competitions, and global institutions.

Blue Ocean Competition
Top 350 of 12,784+ submissions (top 3% internationally), the world's most prestigious student pitch competition
UN Academic Impact
Selected as 1 of 4,000 Millennium Fellows from 60,000+ applicants, connected to UN gender equity frameworks
NCWIT Aspirations in Computing
National Honorable Mention, placing in the top 400 of 3,700+ applicants nationwide
KidsRights International
Top 30 for the International Children's Peace Prize, the world's most prestigious youth prize ($125K+)
SDG Global Citizens Challenge
Top 65 internationally, top 5 for SDG 5 (Gender Equality), validating relevance to global equity goals
ROAR & Penn Engineering
Expert input from ROAR (Philadelphia femtech company) and Penn electrical engineering professors on AI architecture

Market Opportunity

A $120B market with no proactive AI solution.

HavenKey sits at the intersection of wearable devices, mobile safety apps, and emergency-alert systems, a market projected to grow at 11–12% CAGR through 2030. No current solution offers passive, AI-driven detection rather than reactive alerts.

$120B
TAM · Wearable Safety Market
Total addressable market across all personal safety wearables, mobile apps, and emergency alert systems globally.
$32B
SAM · Women 18–35 & Institutions
College students, working women, and university/NGO institutional safety partners in served geographies.
$450–600M
SOM · D2C & NGO Channels
Obtainable via direct-to-consumer and NGO distribution, anchored by the G.I.R.L.S. pilot network across 17 countries.

Business Model

Three revenue streams. Hardware, subscription, institutional.

Hardware Sale

$99

One-time device purchase. Target manufacturing cost ~$35/unit. 55–60% target margin.

Subscription

$4.99/mo

24/7 monitoring, alerts, and incident history. ARPU ~$99 + $20/year. CPA target $15–$25.

Institutional

$1–3K

Per campus deployment via university safety offices, NGOs, and women's shelters. B2B2C channel.

Long-Term Vision

Beyond a product. A movement for women's safety.

HavenKey's vision is not a single device. It's a safety ecosystem that connects with smart cities, scales across institutions, and evolves continuously with the communities it serves.

Global Vision

Every woman safe, empowered, and free from fear, regardless of location, income, or access to emergency services.

Smart City Integration

Connects with smart homes and cities, triggering alarms, locks, and lights. Direct link to emergency services, shelters, and legal aid.

Evolving with Feedback

G.I.R.L.S. members will label real sensor data, shaping model iterations. Turning pilot users into co-designers of the next version.

Global Ecosystem

Wearables, smart technology, and community support: a layered infrastructure for women's safety that scales across institutions and NGOs worldwide.

About the Founder

Divyansha Nashine

I am a sophomore at the University of Pennsylvania in the Jerome Fisher M&T Program, pursuing dual degrees in Mechanical Engineering and Applied Mechanics at Penn Engineering and Economics at the Wharton School.

I co-founded the G.I.R.L.S. Organization at 15, and have since built it into a global STEM movement spanning 30+ U.S. states and 17 countries. I conduct engineering research at Penn's Architected Materials Laboratory and previously at Princeton University, where my findings were published in the Journal of Applied Physics. I am currently an Applied AI Engineer intern at Harness, working on agentic evaluation pipelines.

HavenKey is not a pivot away from that work. It is the direct result of it: four years of listening to girls in my own community describe the same fear, combined with the engineering background to finally build something for it.

LinkedIn G.I.R.L.S. Organization
Divyansha Nashine

Founder, HavenKey & G.I.R.L.S. Organization

Get Involved

No girl should have to choose between her safety and her future.

HavenKey is currently in active development. If you're a researcher, partner, or member of the G.I.R.L.S. community interested in our pilot program, reach out.

Email Divyansha Email the Team

nashine9@engineering.upenn.edu  ·  connect.inspiregirls@gmail.com

G.I.R.L.S. Organization →