Four slides in ten days. Then the data spoke.
🧑🔬 Saikhom Kiba, 34, is a civil engineer for BRO (Border Roads Organisation) maintaining NH-54 from Aizawl toward the Myanmar border. Every June, the monsoon closes his work window for four months. His salary is ₹85,000 per month, with quarterly bonuses if blockages don't exceed eight days. Last June, he faced the monsoon's worst test.
🚨 The problem
In ten days — June 3 to June 14 — four landslides hit the 73-kilometre NH-54 corridor. The first two he cleared on budget, but the third and fourth hit simultaneously: two slopes failed eight hours apart, and Saikhom's three positioned units could not cover both sites. Road blockages stretched to 96 hours. BRO headquarters initiated a contract review. His superior asked the quiet question: how did you know where to position the equipment? Saikhom had guessed.
🚀 How GabFORGE helped
In July, Saikhom attended a disaster-resilience workshop at Aizawl PWD and met Ananya Sharma, a structural engineer from the smart-city mission, who mentioned an AI agent that could predict slope failures using rainfall forecasts, topographic data, and historical records. Over one Thursday evening, they fed the agent three datasets:
- 🔍 Historical slope-failure records from BRO archives: 27 years of landslide data (1998–2025) — location, rainfall, blockage duration.
- 📏 GPS slope measurements: Every 500-metre section's slope angle, measured in the 2024 dry season.
- 📸 Site visit photographs: Seepage stains, exfoliation patterns, and geomorphic markers from three years of surveys.
The agent linked to IMD's Aizawl rainfall API and generated a predicted failure-probability table, colour-coded by section. By July, Saikhom ran the agent backwards over June 2025. Result: on June 12 eve, the three sections that failed had 28%, 31%, and 24% failure probability. He had possessed this information since July of the previous year.
In May 2026, Colonel Rijiju approved Saikhom's renewal with a performance bonus: if monsoon blockage averages fall below 36 hours, Saikhom receives ₹15,000 monthly bonus during the dry season — proof that prediction works where defence could not.
🇮🇳 Why this matters
Border regions like Mizoram have one engineer per twenty-four kilometres of road, three equipment units, and 150+ rainy days annually. They cannot be everywhere. But they can be in the most dangerous places on the most dangerous days, if an agent tells them where to go before the slope fails. That is the boundary between defending and predicting.
The long version has the timeline, Saikhom's pencil-marked maps, his wife's question about nervousness, and the moment the agent showed him he had been guessing all along.