IGAD Hackathon 2026 - Smarter Early Warning, Stronger Communities
Warnings people actually understand.
Kestrel turns raw hazard data into plain-language, hyperlocal warnings - and is the first system in this space that verifies the warning was received and understood, not just sent.
The gap IGAD named
Most early-warning systems stop at “alert issued.”
Information has to be generated, communicated, understood, and turned into action. Generic SMS blasts and technical bulletins routinely fail at the last two steps - and there is no mechanism to check who actually received and comprehended a warning until damage is already done. Kestrel closes that loop, end to end.
Signal
Real weather and hazard readings come in from the ground - no simulated numbers pretending to be live.
Generate
An AI layer turns the raw reading into a short, plain-language, action-first warning - not a meteorological bulletin.
Communicate
The warning goes out across channels people actually use - a real Telegram message, simulated SMS and USSD.
Understand
Recipients confirm with one tap: understood, or need more info. No confirmation, no assumption of safety.
Escalate
When confirmation lags, a second AI pass reads the gap and recommends a concrete fallback - like switching to radio.
A kestrel hunts by hovering perfectly still, watching the ground for the smallest signal of movement - then acts, precisely, the moment it matters.
Who this is for
Built for the officers coordinating the response.
National and regional disaster-management officers and NGOs are the primary users - the decision-makers IGAD serves directly. Community focal points confirm receipt on the ground; the at-risk community itself is represented only as aggregate confirmation data, never individual identity.
Every warning here is generated from a real reading - see it move through the loop.