Consul

AI Emergency Line for Migrants

Comment

CONSUL (777)

An AI Emergency Line for Migrants

🌍 What Inspired This Project

The idea for CONSUL (777) came from a simple but uncomfortable truth:
in moments of crisis, migrants often do not know who to call, what to say, or how to say it.

Many emergency systems assume:

  • stable internet access,
  • fluency in the local language,
  • familiarity with legal or bureaucratic processes,
  • and the emotional capacity to explain a crisis clearly.

In reality, migrants and travelers often face emergencies while:

  • offline or with limited connectivity,
  • using borrowed phones,
  • under stress, fear, or panic,
  • unsure of their rights or nearest embassy contacts.

I was inspired by how emergency numbers like 911 or 999 reduce complexity to one action.
I asked myself:

What if migrants had a single number that didn’t require them to know what to say, who to call, or which forms to fill?

That question became CONSUL (777).

🧠 What I Learned

Building this project taught me several key lessons:

  1. AI is most powerful when it removes cognitive load, not when it adds features.
    The goal was not a “smart chatbot,” but a calm, structured listener that works under pressure.

  2. Emergency systems are workflows, not apps.
    The real value lies in automating what happens after the call: data structuring, routing, notifications, and escalation.

  3. Trust matters more than novelty.
    For a system touching embassies and personal data, clarity and restraint matter more than flashy interfaces.

  4. Offline-first thinking changes architecture decisions.
    Designing for voice calls instead of screens reshaped everything.

🛠️ How I Built the Project

1. Core Concept

CONSUL (777) is an AI-powered emergency hotline that migrants can call from any phone.
Instead of speaking to a human operator, callers speak to an AI agent that:

  • Collects essential information through natural conversation
  • Works even with limited or unstable connectivity
  • Structures the data automatically
  • Sends verified emergency reports to:
    • relevant embassies or consular offices
    • pre-listed emergency contacts

The user does not need to explain what to do.
The system knows.

2. Technical Architecture (High Level)

Input Layer

  • Voice call to short-code number (e.g. 777)
  • Speech-to-text pipeline

AI Layer

  • Google Gemini 3 for:
    • intent detection
    • structured data extraction
    • multilingual understanding
    • calm, guided conversational flow

Mathematically, the extraction step can be seen as:

[ \text{Conversation} \xrightarrow{\text{Gemini}} { \text{Name}, \text{Nationality}, \text{Location}, \text{Emergency Type}, \text{Consent} } ]

Processing Layer

  • Validation and normalization of inputs
  • Location handling via:
    • user-declared location
    • optional device-based location if available

Output Layer

  • Automated dispatch to:
    • embassy systems
    • emergency contacts
  • Logged case reference for follow-up

3. Why Gemini 3

Gemini 3 was critical because it:

  • Handles multilingual voice conversations
  • Maintains context across stressed, unstructured speech
  • Converts free-form dialogue into clean, machine-readable records
  • Allows rapid prototyping without building a rule-based dialogue tree

Without Gemini, the system would either be brittle or slow.

⚠️ Challenges I Faced

1. Location Without Assumptions

A major challenge was handling location when callers might:

  • use borrowed phones,
  • be offline,
  • not want automatic tracking.

The solution was consent-first, declaration-first:

  • Ask the caller to state their location
  • Offer optional location sharing if available
  • Never assume silently

2. Designing for Panic

People in distress do not speak in neat sentences.

I had to design prompts that:

  • are short,
  • avoid legal language,
  • feel calming rather than interrogative.

3. Balancing Privacy and Urgency

Emergency response requires speed, but migrants are rightly sensitive about data sharing.

CONSUL only collects:

  • what is necessary,
  • what the user consents to,
  • and only for the purpose of emergency escalation.

4. Making It Feel “Real”

One risk was building something that felt like a concept, not infrastructure.

To avoid this, I focused on:

  • realistic embassy workflows,
  • credible naming and tone,
  • minimal but clear scope.

🚀 Potential Impact

If deployed, CONSUL could:

  • Reduce response time during migrant emergencies
  • Lower language and knowledge barriers
  • Act as a digital extension of embassies
  • Serve millions of travelers, refugees, and migrant workers globally

This is not just an app.
It is emergency infrastructure powered by AI.

🧭 Closing Thought

CONSUL is built on one belief:

In an emergency, people should not have to be articulate, connected, or informed.
They should only have to ask for help.

AI makes that possible.

Share this project:

Updates