From raw environmental data to life-saving predictions — powered by machine learning.
From Bangkok to Houston, Jakarta to Lagos — communities worldwide face recurrent flooding with minimal early warning. Traditional hydrological models require expensive infrastructure, specialized expertise, and dense sensor networks that most regions simply cannot afford. SurgeShield changes that, turning a handful of measurable environmental signals into an instant, interpretable flood-risk assessment that runs anywhere.
STEP 01
10,000 real-world flood observations, each with 11 environmental features — rainfall, river discharge, elevation, soil type and more.
STEP 02
Three ML algorithms compared head-to-head: Logistic Regression, Random Forest and XGBoost.
STEP 03
The best model — Logistic Regression at —% accuracy — is served through a production REST API on a live VPS, behind HTTPS.
Python
ML pipeline core
Flask
REST ML API
scikit-learn
Model training
XGBoost
Gradient boosting
Next.js
Web frontend
React
UI components
Convex
Database
Clerk
Authentication
Leaflet.js
Interactive maps
Recharts
Data viz
Nginx
Reverse proxy
Let's Encrypt
HTTPS / SSL
Building resilience to natural disasters and reducing the human and economic toll of flooding on urban populations.
Strengthening adaptive capacity to climate-related hazards and natural disasters across vulnerable regions.
SurgeShield demonstrates that AI-powered disaster prediction can be made accessible, interpretable, and deployable anywhere in the world — from developed nations to underserved communities.
ICT University, Cameroon
Faculty of Information and Communication Technology
Department of Software Engineering · Class of 2026