Skip to content
All work
IEEE · acceptedML Systems2025 - 2026Co-author · ML & backend

EcoSphere-AI

Explainable AI for climate-smart housing

View codeRead the paper
AUC
0.92AUC
locations evaluated
14,522locations evaluated
hazards modelled
3hazards modelled

The problem

City planners and families decide where to build and live using static maps and old averages, while climate risk shifts year by year.

What I built

Google Earth Engine pulls multisource geospatial data; BigQuery processes it; an XGBoost classifier scores risk in real time; SHAP explains every prediction; a Gemini assistant answers in plain language.

Highlights

  • Composite Climate Risk Index from flood, heat and drought signals
  • XGBoost tuned on a globally diverse set - 14,522 locations tested
  • Containerized on Google Cloud with a Next.js map interface

How it works

  1. 01Earth Engine pullMultisource geospatial
  2. 02BigQuery processing
  3. 03Risk index buildFlood · heat · drought
  4. 04XGBoost scoring
  5. 05SHAP explanation
  6. 06Gemini answerIn plain language
Next projectTravelSyncAI itinerary planner